feat: 合并上游 Rust 实现，扩展 API/运行时/工具链能力

将 claw-code/rust/crates 的完整实现合并到主 workspace，涵盖 9 个 crate 的更新与 2 个新 crate 的引入。 API 层： - 用原生 Anthropic 客户端（anthropic.rs）替换 claw_provider，新增 prompt cache 减少重复请求开销 - 新增 HTTP 客户端构建器统一代理配置，OpenAI 兼容端增加 DashScope/Qwen 支持与抖动重试 - MessageRequest 扩展 temperature/top_p 等模型调参字段 - SSE 解析器增加 provider 上下文感知的错误信息运行时（~11,000 行新增）： - 新增 bash 命令安全校验、分支锁碰撞检测、配置文件校验 - 新增会话存储与控制面、MCP 生命周期状态机与服务端实现 - 新增权限执行引擎、策略引擎、插件生命周期管理 - 新增 worker 启动编排、任务/定时任务注册表、信任解析器 - 保留 Windows cmd /C fallback 命令/插件/工具： - commands 大幅重写，扩展 sandbox、doctor、plan 等 slash 命令 - plugins 新增 PostToolUseFailure hook 与宽容加载机制 - tools 新增 PDF 提取与 lane 补全工具新增 crate：mock-anthropic-service（测试）、telemetry（遥测）适配 claw-cli/server：ClawApiClient→AnthropicClient 重命名， SlashCommand::parse 返回 Result，移除 session 级 Thinking 变体， TokenUsage/ConversationMessage 补充序列化支持
feat: 添加 Web 前端及服务端 SSE 流式支持，扩展多模型兼容
2026-04-13 14:39:17 +08:00 · 2026-04-10 16:29:27 +08:00
130 changed files with 57270 additions and 4211 deletions
@@ -0,0 +1,5 @@
+{
+  "permissions": {
+    "defaultMode": "dontAsk"
+  }
+}
@@ -0,0 +1,3 @@
+ANTHROPIC_API_KEY="9494feba6f7c45f48c3dfc35a85ffd89.2WUCscxcSp92ETNg"
+ANTHROPIC_BASE_URL="https://open.bigmodel.cn/api/anthropic"
+CLAW_MODEL="glm-5"
@@ -1,3 +1,6 @@
 target/
 .omx/
 .clawd-agents/
+# Claw Code local artifacts
+.claw/
+.claude/
@@ -0,0 +1,113 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+Claw Code is a local coding-agent CLI tool written in safe Rust. It is a clean-room implementation inspired by Claude Code, providing an interactive REPL, one-shot prompts, workspace-aware tools, local agent workflows, and plugin support. The project name and references throughout use "claw" / "Claw Code".
+
+## Build & Run Commands
+
+```bash
+# Build release binary (produces target/release/claw)
+cargo build --release -p claw-cli
+
+# Run from source (interactive REPL)
+cargo run --bin claw --
+
+# Run one-shot prompt
+cargo run --bin claw -- prompt "summarize this workspace"
+
+# Install locally
+cargo install --path crates/claw-cli --locked
+
+# Run the HTTP server binary
+cargo run --bin claw-server
+```
+
+## Verification Commands
+
+```bash
+# Format check
+cargo fmt
+
+# Lint (workspace-level clippy with deny warnings)
+cargo clippy --workspace --all-targets -- -D warnings
+
+# Run all tests
+cargo test --workspace
+
+# Run tests for a specific crate
+cargo test -p api
+cargo test -p runtime
+
+# Run a single test by name
+cargo test -p <crate> -- <test_name>
+
+# Integration tests in crates/api/tests/ use mock TCP servers (no network needed)
+# One smoke test is #[ignore] — run with: cargo test -p api -- --ignored
+```
+
+## Workspace Architecture
+
+Cargo workspace with `resolver = "2"`. All crates live under `crates/`.
+
+### Crate Dependency Graph
+
+```
+claw-cli  ──→  api, runtime, tools, commands, plugins, compat-harness
+server    ──→  api, runtime, tools, plugins, commands
+tools     ──→  api, runtime, plugins
+commands  ──→  runtime, plugins
+api       ──→  runtime
+runtime   ──→  lsp, plugins
+plugins   ──→  (standalone, serde only)
+lsp       ──→  (standalone)
+compat-harness ──→  commands, tools, runtime
+```
+
+### Core Crates
+
+- **`claw-cli`** — User-facing binary (`claw`). REPL loop with markdown rendering (pulldown-cmark + syntect), argument parsing, OAuth flow. Entry point: `crates/claw-cli/src/main.rs`.
+
+- **`runtime`** — Session management, conversation runtime, permissions, system prompt construction, context compaction, MCP stdio management, and hook execution. Key types: `Session` (versioned message history), `ConversationRuntime<C, T>` (generic over `ApiClient` + `ToolExecutor` traits), `PermissionMode`, `McpServerManager`, `HookRunner`.
+
+- **`api`** — HTTP client for LLM providers with SSE streaming. `ClawApiClient` (Anthropic-compatible), `OpenAiCompatClient`, and `Provider` trait. `ProviderKind` enum distinguishes ClawApi, Xai, OpenAi. Request/response types: `MessageRequest`, `StreamEvent`, `ToolDefinition`.
+
+- **`tools`** — Built-in tool definitions and dispatch. `GlobalToolRegistry` is a lazy-static singleton. Tools: Read, Write, Edit, Glob, Grep, Bash, LSP, Task*, Cron*, Worktree*. Each tool has a `ToolSpec` with JSON schema.
+
+- **`commands`** — Slash command registry and handlers (`/help`, `/config`, `/compact`, `/resume`, `/plugins`, `/agents`, `/doctor`, etc.). `SlashCommandSpec` defines each command's name, aliases, description, and category.
+
+- **`plugins`** — Plugin discovery and lifecycle. `PluginManager` loads builtin, bundled, and external (from `~/.claw/plugins/`) plugins. Plugins can provide additional tools via `PluginTool`.
+
+- **`server`** — Axum-based HTTP server (`claw-server`). REST endpoints for session CRUD + SSE event streaming. `AppState` holds shared session store.
+
+- **`lsp`** — Language Server Protocol types and process management for code intelligence features.
+
+- **`compat-harness`** — Extracts command/tool/bootstrap-plan manifests from upstream TypeScript source files (for compatibility tracking). Uses `CLAUDE_CODE_UPSTREAM` env var to locate the upstream repo.
+
+## Key Architectural Patterns
+
+- **Trait-based abstraction**: `ApiClient`, `ToolExecutor`, `Provider` traits enable swappable implementations. `ConversationRuntime` is generic over client and executor.
+- **Static registries**: `GlobalToolRegistry` and slash command specs use lazy-static initialization with compile-time definitions.
+- **SSE streaming**: API responses stream through `MessageStream` (async iterator) to the terminal renderer or server SSE endpoints.
+- **Permission model**: `PermissionMode` enum — ReadOnly, WorkspaceWrite, DangerFullAccess. Configurable via `.claw.json` (`permissions.defaultMode`).
+- **Hook system**: Pre/post tool execution hooks via `HookRunner` in the runtime crate.
+
+## Configuration & Environment
+
+- `.claw.json` — Project-level config (permissions, etc.)
+- `.claw/` — Project directory for hooks, plugins, local settings
+- `~/.claw/` — User-level config directory
+- `.env` — API keys (gitignored): `ANTHROPIC_API_KEY`, `ANTHROPIC_BASE_URL`, `XAI_API_KEY`, `XAI_BASE_URL`
+- `CLAW.md` — Workspace instructions loaded into the system prompt (analogous to CLAUDE.md but for claw itself)
+
+## Lint Rules
+
+- `unsafe_code` is **forbidden** at workspace level
+- Clippy `all` + `pedantic` lints are warnings; some pedantic lints are allowed (`module_name_repetitions`, `missing_panics_doc`, `missing_errors_doc`)
+- CI runs: `cargo check --workspace`, `cargo test --workspace`, `cargo build --release` on Ubuntu and macOS
+
+## Language
+
+Code comments, commit messages, and documentation are primarily in Chinese (中文). UI strings and exported symbol names are in English.
@@ -0,0 +1,15 @@
+# CLAW.md
+
+This file provides guidance to Claw Code (clawcode.dev) when working with code in this repository.
+
+## Detected stack
+- Languages: Rust.
+- Frameworks: none detected from the supported starter markers.
+
+## Verification
+- Run Rust verification from the repo root: `cargo fmt`, `cargo clippy --workspace --all-targets -- -D warnings`, `cargo test --workspace`
+
+## Working agreement
+- Prefer small, reviewable changes and keep generated bootstrap files aligned with actual repo workflows.
+- Keep shared defaults in `.claw.json`; reserve `.claw/settings.local.json` for machine-local overrides.
+- Do not overwrite existing `CLAW.md` content automatically; update it intentionally when repo workflows change.
@@ -17,6 +17,15 @@ dependencies = [
 "memchr",
 ]

+[[package]]
+name = "android_system_properties"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "819e7219dbd41043ac279b19830f2efc897156490d7fd6ea916720117ee66311"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "api"
 version = "0.1.0"
@@ -25,6 +34,7 @@ dependencies = [
 "runtime",
 "serde",
 "serde_json",
+ "telemetry",
 "tokio",
 ]

@@ -56,6 +66,12 @@ version = "1.1.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1505bd5d3d116872e7271a6d4e16d81d0c8570876c8de68093a09ac269d8aac0"

+[[package]]
+name = "autocfg"
+version = "1.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c08606f8c3cbf4ce6ec8e28fb0014a2c086708fe954eaa885384a6165172e7e8"
+
 [[package]]
 name = "axum"
 version = "0.8.8"
@@ -178,6 +194,19 @@ version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "613afe47fcd5fac7ccf1db93babcb082c5994d996f20b8b159f2ad1658eb5724"

+[[package]]
+name = "chrono"
+version = "0.4.44"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c673075a2e0e5f4a1dde27ce9dee1ea4558c7ffe648f576438a20ca1d2acc4b0"
+dependencies = [
+ "iana-time-zone",
+ "js-sys",
+ "num-traits",
+ "wasm-bindgen",
+ "windows-link",
+]
+
 [[package]]
 name = "claw-cli"
 version = "0.1.0"
@@ -186,6 +215,7 @@ dependencies = [
 "commands",
 "compat-harness",
 "crossterm",
+ "dotenvy",
 "plugins",
 "pulldown-cmark",
 "runtime",
@@ -223,6 +253,12 @@ dependencies = [
 "tools",
 ]

+[[package]]
+name = "core-foundation-sys"
+version = "0.8.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "773648b94d0e5d620f64f280777445740e61fe701025087ec8b57f45c791888b"
+
 [[package]]
 name = "cpufeatures"
 version = "0.2.17"
@@ -306,12 +342,30 @@ dependencies = [
 "syn",
 ]

+[[package]]
+name = "dotenvy"
+version = "0.15.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1aaf95b3e5c8f23aa320147307562d361db0ae0d51242340f558153b4eb2439b"
+
+[[package]]
+name = "either"
+version = "1.15.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "48c757948c5ede0e46177b7add2e67155f70e33c07fea8284df6576da70b3719"
+
 [[package]]
 name = "endian-type"
 version = "0.1.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c34f04666d835ff5d62e058c3995147c06f42fe86ff053337632bca83e42702d"

+[[package]]
+name = "env_home"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c7f84e12ccf0a7ddc17a6c41c93326024c42920d7ee630d04950e6926645c0fe"
+
 [[package]]
 name = "equivalent"
 version = "1.0.2"
@@ -619,6 +673,30 @@ dependencies = [
 "tracing",
 ]

+[[package]]
+name = "iana-time-zone"
+version = "0.1.65"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e31bc9ad994ba00e440a8aa5c9ef0ec67d5cb5e5cb0cc7f8b744a35b389cc470"
+dependencies = [
+ "android_system_properties",
+ "core-foundation-sys",
+ "iana-time-zone-haiku",
+ "js-sys",
+ "log",
+ "wasm-bindgen",
+ "windows-core",
+]
+
+[[package]]
+name = "iana-time-zone-haiku"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f31827a206f56af32e590ba56d5d2d085f558508192593743f16b2306495269f"
+dependencies = [
+ "cc",
+]
+
 [[package]]
 name = "icu_collections"
 version = "2.1.1"
@@ -880,6 +958,15 @@ dependencies = [
 "windows-sys 0.61.2",
 ]

+[[package]]
+name = "mock-anthropic-service"
+version = "0.1.0"
+dependencies = [
+ "api",
+ "serde_json",
+ "tokio",
+]
+
 [[package]]
 name = "nibble_vec"
 version = "0.1.0"
@@ -907,6 +994,15 @@ version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c6673768db2d862beb9b39a78fdcb1a69439615d5794a1be50caa9bc92c81967"

+[[package]]
+name = "num-traits"
+version = "0.2.19"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "071dfc062690e90b734c0b2273ce72ad0ffa95f0c74596bc250dcfd960262841"
+dependencies = [
+ "autocfg",
+]
+
 [[package]]
 name = "once_cell"
 version = "1.21.4"
@@ -1266,14 +1362,15 @@ name = "runtime"
 version = "0.1.0"
 dependencies = [
 "glob",
- "lsp",
 "plugins",
 "regex",
 "serde",
 "serde_json",
 "sha2",
+ "telemetry",
 "tokio",
 "walkdir",
+ "which",
 ]

 [[package]]
@@ -1357,9 +1454,12 @@ dependencies = [
 "commands",
 "compat-harness",
 "crossterm",
+ "mock-anthropic-service",
 "plugins",
 "pulldown-cmark",
 "runtime",
+ "rustyline",
+ "serde",
 "serde_json",
 "syntect",
 "tokio",
@@ -1490,13 +1590,21 @@ dependencies = [
 name = "server"
 version = "0.1.0"
 dependencies = [
+ "api",
 "async-stream",
 "axum",
+ "chrono",
+ "commands",
+ "dotenvy",
+ "plugins",
 "reqwest",
 "runtime",
 "serde",
 "serde_json",
 "tokio",
+ "tools",
+ "tower",
+ "tower-http",
 ]

 [[package]]
@@ -1639,6 +1747,14 @@ dependencies = [
 "yaml-rust",
 ]

+[[package]]
+name = "telemetry"
+version = "0.1.0"
+dependencies = [
+ "serde",
+ "serde_json",
+]
+
 [[package]]
 name = "thiserror"
 version = "2.0.18"
@@ -1770,6 +1886,8 @@ name = "tools"
 version = "0.1.0"
 dependencies = [
 "api",
+ "commands",
+ "flate2",
 "plugins",
 "reqwest",
 "runtime",
@@ -2047,6 +2165,18 @@ dependencies = [
 "rustls-pki-types",
 ]

+[[package]]
+name = "which"
+version = "7.0.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "24d643ce3fd3e5b54854602a080f34fb10ab75e0b813ee32d00ca2b44fa74762"
+dependencies = [
+ "either",
+ "env_home",
+ "rustix 1.1.4",
+ "winsafe",
+]
+
 [[package]]
 name = "winapi"
 version = "0.3.9"
@@ -2078,12 +2208,65 @@ version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"

+[[package]]
+name = "windows-core"
+version = "0.62.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b8e83a14d34d0623b51dce9581199302a221863196a1dde71a7663a4c2be9deb"
+dependencies = [
+ "windows-implement",
+ "windows-interface",
+ "windows-link",
+ "windows-result",
+ "windows-strings",
+]
+
+[[package]]
+name = "windows-implement"
+version = "0.60.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "053e2e040ab57b9dc951b72c264860db7eb3b0200ba345b4e4c3b14f67855ddf"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "windows-interface"
+version = "0.59.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3f316c4a2570ba26bbec722032c4099d8c8bc095efccdc15688708623367e358"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "windows-link"
 version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f0805222e57f7521d6a62e36fa9163bc891acd422f971defe97d64e70d0a4fe5"

+[[package]]
+name = "windows-result"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7781fa89eaf60850ac3d2da7af8e5242a5ea78d1a11c49bf2910bb5a73853eb5"
+dependencies = [
+ "windows-link",
+]
+
+[[package]]
+name = "windows-strings"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7837d08f69c77cf6b07689544538e017c1bfcf57e34b4c0ff58e6c2cd3b37091"
+dependencies = [
+ "windows-link",
+]
+
 [[package]]
 name = "windows-sys"
 version = "0.52.0"
@@ -2249,6 +2432,12 @@ version = "0.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d6bbff5f0aada427a1e5a6da5f1f98158182f26556f345ac9e04d36d0ebed650"

+[[package]]
+name = "winsafe"
+version = "0.0.19"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d135d17ab770252ad95e9a872d365cf3090e3be864a34ab46f48555993efc904"
+
 [[package]]
 name = "wit-bindgen"
 version = "0.51.0"
@@ -21,3 +21,35 @@ pedantic = { level = "warn", priority = -1 }
 module_name_repetitions = "allow"
 missing_panics_doc = "allow"
 missing_errors_doc = "allow"
+uninlined_format_args = "allow"
+map_unwrap_or = "allow"
+doc_markdown = "allow"
+redundant_pattern_matching = "allow"
+items_after_statements = "allow"
+must_use_candidate = "allow"
+io_other_error = "allow"
+implicit_clone = "allow"
+redundant_closure_for_method_calls = "allow"
+unnecessary_map_or = "allow"
+manual_let_else = "allow"
+format_push_string = "allow"
+match_same_arms = "allow"
+similar_names = "allow"
+needless_continue = "allow"
+assigning_clones = "allow"
+cast_possible_truncation = "allow"
+cast_possible_wrap = "allow"
+cast_sign_loss = "allow"
+cmp_owned = "allow"
+collapsible_if = "allow"
+too_many_lines = "allow"
+wildcard_in_or_patterns = "allow"
+explicit_counter_loop = "allow"
+manual_repeat_n = "allow"
+manual_str_repeat = "allow"
+needless_borrow = "allow"
+needless_pass_by_value = "allow"
+single_match_else = "allow"
+too_many_arguments = "allow"
+unnested_or_patterns = "allow"
+unused_self = "allow"
@@ -19,12 +19,11 @@ Rust 工作区是当前主要的产品界面。`claw` 二进制文件在单个
 - Cargo
 - 你想使用的模型的提供商凭据

-### 身份验证
-
-兼容 Anthropic 的模型：
+你可以通过环境变量或在项目根目录创建 **`.env`** 文件来配置 API 密钥：

+**配置 Claude (推荐):**
 ```bash
-export ANTHROPIC_API_KEY="..."
+ANTHROPIC_API_KEY="..."
 # 使用兼容的端点时可选
 export ANTHROPIC_BASE_URL="https://api.anthropic.com"
 ```
@@ -10,6 +10,7 @@ reqwest = { version = "0.12", default-features = false, features = ["json", "rus
 runtime = { path = "../runtime" }
 serde = { version = "1", features = ["derive"] }
 serde_json.workspace = true
+telemetry = { path = "../telemetry" }
 tokio = { version = "1", features = ["io-util", "macros", "net", "rt-multi-thread", "time"] }

 [lints]
@@ -0,0 +1,53 @@
+# API 模块 (api)
+
+本模块提供了与大型语言模型 (LLM) 服务提供商（主要是 Anthropic 的 Claude 和兼容 OpenAI 的服务）进行交互的高层抽象和客户端。
+
+## 概览
+
+`api` 模块负责以下职责：
+- 标准化与不同 AI 提供商的通信。
+- 通过服务器发送事件 (SSE) 处理流式响应。
+- 管理身份验证源（API 密钥、OAuth 令牌）。
+- 提供消息、工具和使用情况跟踪的共享数据结构。
+
+## 关键特性
+
+- **提供商抽象 (Provider Abstraction)**：支持多种 AI 后端，包括：
+  - `ClawApiClient`: Claude 模型的主要提供商。
+  - `OpenAiCompatClient`: 支持兼容 OpenAI 的 API（如本地模型、专门的提供商）。
+- **流式支持 (Streaming Support)**：健壮的 SSE 解析实现 (`SseParser`)，用于处理实时的内容生成。
+- **工具集成 (Tool Integration)**：为 `ToolDefinition`、`ToolChoice` 和 `ToolResultContentBlock` 提供强类型定义，支持智能代理 (Agentic) 工作流。
+- **身份验证管理 (Auth Management)**：用于解析启动身份验证源和管理 OAuth 令牌的实用工具。
+- **模型智能 (Model Intelligence)**：解析模型别名和计算最大标记 (Token) 限制的元数据及辅助函数。
+
+## 实现逻辑
+
+### 核心模块
+
+- **`client.rs`**: 定义了 `ProviderClient` 特性 (Trait) 和基础客户端逻辑。它使用 `reqwest` 处理 HTTP 请求，并管理消息流的生命周期。
+- **`types.rs`**: 包含 API 的核心数据模型，如 `InputMessage`、`OutputContentBlock` 以及 `MessageRequest`/`MessageResponse`。
+- **`sse.rs`**: 实现了一个状态化的 SSE 解析器，能够处理分段的数据块并发出类型化的 `StreamEvent`。
+- **`providers/`**: 包含针对不同 LLM 端点的特定逻辑，将它们的独特格式映射到本模块使用的共享类型。
+
+### 数据流
+
+1. 构建包含模型详情、消息和工具定义的 `MessageRequest`。
+2. `ApiClient` 将此请求转换为提供商特定的 HTTP 请求。
+3. 如果启用了流式传输，客户端返回一个 `MessageStream`，该流使用 `SseParser` 来产生 `StreamEvent`。
+4. 最终响应包含用于跟踪 Token 消耗的 `Usage` 信息。
+
+## 使用示例
+
+```rust
+use api::{ApiClient, MessageRequest, InputMessage};
+
+// 示例初始化（已简化）
+let client = ApiClient::new(auth_source);
+let request = MessageRequest {
+    model: "claude-3-5-sonnet-20241022".to_string(),
+    messages: vec![InputMessage::user("你好，世界！")],
+    ..Default::default()
+};
+
+let stream = client.create_message_stream(request).await?;
+```
@@ -1,70 +1,91 @@
 use crate::error::ApiError;
-use crate::providers::claw_provider::{self, AuthSource, ClawApiClient};
+use crate::prompt_cache::{PromptCache, PromptCacheRecord, PromptCacheStats};
+use crate::providers::anthropic::{self, AnthropicClient, AuthSource};
 use crate::providers::openai_compat::{self, OpenAiCompatClient, OpenAiCompatConfig};
-use crate::providers::{self, Provider, ProviderKind};
+use crate::providers::{self, ProviderKind};
 use crate::types::{MessageRequest, MessageResponse, StreamEvent};

-async fn send_via_provider<P: Provider>(
-    provider: &P,
-    request: &MessageRequest,
-) -> Result<MessageResponse, ApiError> {
-    provider.send_message(request).await
-}
-
-async fn stream_via_provider<P: Provider>(
-    provider: &P,
-    request: &MessageRequest,
-) -> Result<P::Stream, ApiError> {
-    provider.stream_message(request).await
-}
-
+#[allow(clippy::large_enum_variant)]
 #[derive(Debug, Clone)]
 pub enum ProviderClient {
-    ClawApi(ClawApiClient),
+    Anthropic(AnthropicClient),
    Xai(OpenAiCompatClient),
    OpenAi(OpenAiCompatClient),
 }

 impl ProviderClient {
    pub fn from_model(model: &str) -> Result<Self, ApiError> {
-        Self::from_model_with_default_auth(model, None)
+        Self::from_model_with_anthropic_auth(model, None)
    }

-    pub fn from_model_with_default_auth(
+    pub fn from_model_with_anthropic_auth(
        model: &str,
-        default_auth: Option<AuthSource>,
+        anthropic_auth: Option<AuthSource>,
    ) -> Result<Self, ApiError> {
        let resolved_model = providers::resolve_model_alias(model);
        match providers::detect_provider_kind(&resolved_model) {
-            ProviderKind::ClawApi => Ok(Self::ClawApi(match default_auth {
-                Some(auth) => ClawApiClient::from_auth(auth),
-                None => ClawApiClient::from_env()?,
+            ProviderKind::Anthropic => Ok(Self::Anthropic(match anthropic_auth {
+                Some(auth) => AnthropicClient::from_auth(auth),
+                None => AnthropicClient::from_env()?,
            })),
            ProviderKind::Xai => Ok(Self::Xai(OpenAiCompatClient::from_env(
                OpenAiCompatConfig::xai(),
            )?)),
-            ProviderKind::OpenAi => Ok(Self::OpenAi(OpenAiCompatClient::from_env(
-                OpenAiCompatConfig::openai(),
-            )?)),
+            ProviderKind::OpenAi => {
+                // DashScope models (qwen-*) also return ProviderKind::OpenAi because they
+                // speak the OpenAI wire format, but they need the DashScope config which
+                // reads DASHSCOPE_API_KEY and points at dashscope.aliyuncs.com.
+                let config = match providers::metadata_for_model(&resolved_model) {
+                    Some(meta) if meta.auth_env == "DASHSCOPE_API_KEY" => {
+                        OpenAiCompatConfig::dashscope()
+                    }
+                    _ => OpenAiCompatConfig::openai(),
+                };
+                Ok(Self::OpenAi(OpenAiCompatClient::from_env(config)?))
+            }
        }
    }

    #[must_use]
    pub const fn provider_kind(&self) -> ProviderKind {
        match self {
-            Self::ClawApi(_) => ProviderKind::ClawApi,
+            Self::Anthropic(_) => ProviderKind::Anthropic,
            Self::Xai(_) => ProviderKind::Xai,
            Self::OpenAi(_) => ProviderKind::OpenAi,
        }
    }

+    #[must_use]
+    pub fn with_prompt_cache(self, prompt_cache: PromptCache) -> Self {
+        match self {
+            Self::Anthropic(client) => Self::Anthropic(client.with_prompt_cache(prompt_cache)),
+            other => other,
+        }
+    }
+
+    #[must_use]
+    pub fn prompt_cache_stats(&self) -> Option<PromptCacheStats> {
+        match self {
+            Self::Anthropic(client) => client.prompt_cache_stats(),
+            Self::Xai(_) | Self::OpenAi(_) => None,
+        }
+    }
+
+    #[must_use]
+    pub fn take_last_prompt_cache_record(&self) -> Option<PromptCacheRecord> {
+        match self {
+            Self::Anthropic(client) => client.take_last_prompt_cache_record(),
+            Self::Xai(_) | Self::OpenAi(_) => None,
+        }
+    }
+
    pub async fn send_message(
        &self,
        request: &MessageRequest,
    ) -> Result<MessageResponse, ApiError> {
        match self {
-            Self::ClawApi(client) => send_via_provider(client, request).await,
-            Self::Xai(client) | Self::OpenAi(client) => send_via_provider(client, request).await,
+            Self::Anthropic(client) => client.send_message(request).await,
+            Self::Xai(client) | Self::OpenAi(client) => client.send_message(request).await,
        }
    }

@@ -73,10 +94,12 @@ impl ProviderClient {
        request: &MessageRequest,
    ) -> Result<MessageStream, ApiError> {
        match self {
-            Self::ClawApi(client) => stream_via_provider(client, request)
+            Self::Anthropic(client) => client
+                .stream_message(request)
                .await
-                .map(MessageStream::ClawApi),
-            Self::Xai(client) | Self::OpenAi(client) => stream_via_provider(client, request)
+                .map(MessageStream::Anthropic),
+            Self::Xai(client) | Self::OpenAi(client) => client
+                .stream_message(request)
                .await
                .map(MessageStream::OpenAiCompat),
        }
@@ -85,7 +108,7 @@ impl ProviderClient {

 #[derive(Debug)]
 pub enum MessageStream {
-    ClawApi(claw_provider::MessageStream),
+    Anthropic(anthropic::MessageStream),
    OpenAiCompat(openai_compat::MessageStream),
 }

@@ -93,25 +116,25 @@ impl MessageStream {
    #[must_use]
    pub fn request_id(&self) -> Option<&str> {
        match self {
-            Self::ClawApi(stream) => stream.request_id(),
+            Self::Anthropic(stream) => stream.request_id(),
            Self::OpenAiCompat(stream) => stream.request_id(),
        }
    }

    pub async fn next_event(&mut self) -> Result<Option<StreamEvent>, ApiError> {
        match self {
-            Self::ClawApi(stream) => stream.next_event().await,
+            Self::Anthropic(stream) => stream.next_event().await,
            Self::OpenAiCompat(stream) => stream.next_event().await,
        }
    }
 }

-pub use claw_provider::{
+pub use anthropic::{
    oauth_token_is_expired, resolve_saved_oauth_token, resolve_startup_auth_source, OAuthTokenSet,
 };
 #[must_use]
 pub fn read_base_url() -> String {
-    claw_provider::read_base_url()
+    anthropic::read_base_url()
 }

 #[must_use]
@@ -121,8 +144,21 @@ pub fn read_xai_base_url() -> String {

 #[cfg(test)]
 mod tests {
+    use std::sync::{Mutex, OnceLock};
+
+    use super::ProviderClient;
    use crate::providers::{detect_provider_kind, resolve_model_alias, ProviderKind};

+    /// Serializes every test in this module that mutates process-wide
+    /// environment variables so concurrent test threads cannot observe
+    /// each other's partially-applied state.
+    fn env_lock() -> std::sync::MutexGuard<'static, ()> {
+        static LOCK: OnceLock<Mutex<()>> = OnceLock::new();
+        LOCK.get_or_init(|| Mutex::new(()))
+            .lock()
+            .unwrap_or_else(std::sync::PoisonError::into_inner)
+    }
+
    #[test]
    fn resolves_existing_and_grok_aliases() {
        assert_eq!(resolve_model_alias("opus"), "claude-opus-4-6");
@@ -135,7 +171,71 @@ mod tests {
        assert_eq!(detect_provider_kind("grok-3"), ProviderKind::Xai);
        assert_eq!(
            detect_provider_kind("claude-sonnet-4-6"),
-            ProviderKind::ClawApi
+            ProviderKind::Anthropic
        );
    }
+
+    /// Snapshot-restore guard for a single environment variable. Mirrors
+    /// the pattern used in `providers/mod.rs` tests: captures the original
+    /// value on construction, applies the override, and restores on drop so
+    /// tests leave the process env untouched even when they panic.
+    struct EnvVarGuard {
+        key: &'static str,
+        original: Option<std::ffi::OsString>,
+    }
+
+    impl EnvVarGuard {
+        fn set(key: &'static str, value: Option<&str>) -> Self {
+            let original = std::env::var_os(key);
+            match value {
+                Some(value) => std::env::set_var(key, value),
+                None => std::env::remove_var(key),
+            }
+            Self { key, original }
+        }
+    }
+
+    impl Drop for EnvVarGuard {
+        fn drop(&mut self) {
+            match self.original.take() {
+                Some(value) => std::env::set_var(self.key, value),
+                None => std::env::remove_var(self.key),
+            }
+        }
+    }
+
+    #[test]
+    fn dashscope_model_uses_dashscope_config_not_openai() {
+        // Regression: qwen-plus was being routed to OpenAiCompatConfig::openai()
+        // which reads OPENAI_API_KEY and points at api.openai.com, when it should
+        // use OpenAiCompatConfig::dashscope() which reads DASHSCOPE_API_KEY and
+        // points at dashscope.aliyuncs.com.
+        let _lock = env_lock();
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", Some("test-dashscope-key"));
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", None);
+
+        let client = ProviderClient::from_model("qwen-plus");
+
+        // Must succeed (not fail with "missing OPENAI_API_KEY")
+        assert!(
+            client.is_ok(),
+            "qwen-plus with DASHSCOPE_API_KEY set should build successfully, got: {:?}",
+            client.err()
+        );
+
+        // Verify it's the OpenAi variant pointed at the DashScope base URL.
+        match client.unwrap() {
+            ProviderClient::OpenAi(openai_client) => {
+                assert!(
+                    openai_client.base_url().contains("dashscope.aliyuncs.com"),
+                    "qwen-plus should route to DashScope base URL (contains 'dashscope.aliyuncs.com'), got: {}",
+                    openai_client.base_url()
+                );
+            }
+            other => panic!(
+                "Expected ProviderClient::OpenAi for qwen-plus, got: {:?}",
+                other
+            ),
+        }
+    }
 }
@@ -2,22 +2,55 @@ use std::env::VarError;
 use std::fmt::{Display, Formatter};
 use std::time::Duration;

+const GENERIC_FATAL_WRAPPER_MARKERS: &[&str] = &[
+    "something went wrong while processing your request",
+    "please try again, or use /new to start a fresh session",
+];
+
+const CONTEXT_WINDOW_ERROR_MARKERS: &[&str] = &[
+    "maximum context length",
+    "context window",
+    "context length",
+    "too many tokens",
+    "prompt is too long",
+    "input is too long",
+    "request is too large",
+];
+
 #[derive(Debug)]
 pub enum ApiError {
    MissingCredentials {
        provider: &'static str,
        env_vars: &'static [&'static str],
+        /// Optional, runtime-computed hint appended to the error Display
+        /// output. Populated when the provider resolver can infer what the
+        /// user probably intended (e.g. an OpenAI key is set but Anthropic
+        /// was selected because no Anthropic credentials exist).
+        hint: Option<String>,
+    },
+    ContextWindowExceeded {
+        model: String,
+        estimated_input_tokens: u32,
+        requested_output_tokens: u32,
+        estimated_total_tokens: u32,
+        context_window_tokens: u32,
    },
    ExpiredOAuthToken,
    Auth(String),
    InvalidApiKeyEnv(VarError),
    Http(reqwest::Error),
    Io(std::io::Error),
-    Json(serde_json::Error),
+    Json {
+        provider: String,
+        model: String,
+        body_snippet: String,
+        source: serde_json::Error,
+    },
    Api {
        status: reqwest::StatusCode,
        error_type: Option<String>,
        message: Option<String>,
+        request_id: Option<String>,
        body: String,
        retryable: bool,
    },
@@ -38,7 +71,48 @@ impl ApiError {
        provider: &'static str,
        env_vars: &'static [&'static str],
    ) -> Self {
-        Self::MissingCredentials { provider, env_vars }
+        Self::MissingCredentials {
+            provider,
+            env_vars,
+            hint: None,
+        }
+    }
+
+    /// Build a `MissingCredentials` error carrying an extra, runtime-computed
+    /// hint string that the Display impl appends after the canonical "missing
+    /// <provider> credentials" message. Used by the provider resolver to
+    /// suggest the likely fix when the user has credentials for a different
+    /// provider already in the environment.
+    #[must_use]
+    pub fn missing_credentials_with_hint(
+        provider: &'static str,
+        env_vars: &'static [&'static str],
+        hint: impl Into<String>,
+    ) -> Self {
+        Self::MissingCredentials {
+            provider,
+            env_vars,
+            hint: Some(hint.into()),
+        }
+    }
+
+    /// Build a `Self::Json` enriched with the provider name, the model that
+    /// was requested, and the first 200 characters of the raw response body so
+    /// that callers can diagnose deserialization failures without re-running
+    /// the request.
+    #[must_use]
+    pub fn json_deserialize(
+        provider: impl Into<String>,
+        model: impl Into<String>,
+        body: &str,
+        source: serde_json::Error,
+    ) -> Self {
+        Self::Json {
+            provider: provider.into(),
+            model: model.into(),
+            body_snippet: truncate_body_snippet(body, 200),
+            source,
+        }
    }

    #[must_use]
@@ -48,11 +122,106 @@ impl ApiError {
            Self::Api { retryable, .. } => *retryable,
            Self::RetriesExhausted { last_error, .. } => last_error.is_retryable(),
            Self::MissingCredentials { .. }
+            | Self::ContextWindowExceeded { .. }
            | Self::ExpiredOAuthToken
            | Self::Auth(_)
            | Self::InvalidApiKeyEnv(_)
            | Self::Io(_)
-            | Self::Json(_)
+            | Self::Json { .. }
+            | Self::InvalidSseFrame(_)
+            | Self::BackoffOverflow { .. } => false,
+        }
+    }
+
+    #[must_use]
+    pub fn request_id(&self) -> Option<&str> {
+        match self {
+            Self::Api { request_id, .. } => request_id.as_deref(),
+            Self::RetriesExhausted { last_error, .. } => last_error.request_id(),
+            Self::MissingCredentials { .. }
+            | Self::ContextWindowExceeded { .. }
+            | Self::ExpiredOAuthToken
+            | Self::Auth(_)
+            | Self::InvalidApiKeyEnv(_)
+            | Self::Http(_)
+            | Self::Io(_)
+            | Self::Json { .. }
+            | Self::InvalidSseFrame(_)
+            | Self::BackoffOverflow { .. } => None,
+        }
+    }
+
+    #[must_use]
+    pub fn safe_failure_class(&self) -> &'static str {
+        match self {
+            Self::RetriesExhausted { .. } if self.is_context_window_failure() => "context_window",
+            Self::RetriesExhausted { .. } if self.is_generic_fatal_wrapper() => {
+                "provider_retry_exhausted"
+            }
+            Self::RetriesExhausted { last_error, .. } => last_error.safe_failure_class(),
+            Self::MissingCredentials { .. } | Self::ExpiredOAuthToken | Self::Auth(_) => {
+                "provider_auth"
+            }
+            Self::Api { status, .. } if matches!(status.as_u16(), 401 | 403) => "provider_auth",
+            Self::ContextWindowExceeded { .. } => "context_window",
+            Self::Api { .. } if self.is_context_window_failure() => "context_window",
+            Self::Api { status, .. } if status.as_u16() == 429 => "provider_rate_limit",
+            Self::Api { .. } if self.is_generic_fatal_wrapper() => "provider_internal",
+            Self::Api { .. } => "provider_error",
+            Self::Http(_) | Self::InvalidSseFrame(_) | Self::BackoffOverflow { .. } => {
+                "provider_transport"
+            }
+            Self::InvalidApiKeyEnv(_) | Self::Io(_) | Self::Json { .. } => "runtime_io",
+        }
+    }
+
+    #[must_use]
+    pub fn is_generic_fatal_wrapper(&self) -> bool {
+        match self {
+            Self::Api { message, body, .. } => {
+                message
+                    .as_deref()
+                    .is_some_and(looks_like_generic_fatal_wrapper)
+                    || looks_like_generic_fatal_wrapper(body)
+            }
+            Self::RetriesExhausted { last_error, .. } => last_error.is_generic_fatal_wrapper(),
+            Self::MissingCredentials { .. }
+            | Self::ContextWindowExceeded { .. }
+            | Self::ExpiredOAuthToken
+            | Self::Auth(_)
+            | Self::InvalidApiKeyEnv(_)
+            | Self::Http(_)
+            | Self::Io(_)
+            | Self::Json { .. }
+            | Self::InvalidSseFrame(_)
+            | Self::BackoffOverflow { .. } => false,
+        }
+    }
+
+    #[must_use]
+    pub fn is_context_window_failure(&self) -> bool {
+        match self {
+            Self::ContextWindowExceeded { .. } => true,
+            Self::Api {
+                status,
+                message,
+                body,
+                ..
+            } => {
+                matches!(status.as_u16(), 400 | 413 | 422)
+                    && (message
+                        .as_deref()
+                        .is_some_and(looks_like_context_window_error)
+                        || looks_like_context_window_error(body))
+            }
+            Self::RetriesExhausted { last_error, .. } => last_error.is_context_window_failure(),
+            Self::MissingCredentials { .. }
+            | Self::ExpiredOAuthToken
+            | Self::Auth(_)
+            | Self::InvalidApiKeyEnv(_)
+            | Self::Http(_)
+            | Self::Io(_)
+            | Self::Json { .. }
            | Self::InvalidSseFrame(_)
            | Self::BackoffOverflow { .. } => false,
        }
@@ -62,10 +231,43 @@ impl ApiError {
 impl Display for ApiError {
    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
        match self {
-            Self::MissingCredentials { provider, env_vars } => write!(
+            Self::MissingCredentials {
+                provider,
+                env_vars,
+                hint,
+            } => {
+                write!(
+                    f,
+                    "missing {provider} credentials; export {} before calling the {provider} API",
+                    env_vars.join(" or ")
+                )?;
+                if cfg!(target_os = "windows") {
+                    if let Some(primary) = env_vars.first() {
+                        write!(
+                            f,
+                            " (on Windows, environment variables set in PowerShell only persist for the current session; use `setx {primary} <value>` to make it permanent, then open a new terminal, or place a `.env` file containing `{primary}=<value>` in the current working directory)"
+                        )?;
+                    } else {
+                        write!(
+                            f,
+                            " (on Windows, environment variables set in PowerShell only persist for the current session; use `setx` to make them permanent, then open a new terminal, or place a `.env` file in the current working directory)"
+                        )?;
+                    }
+                }
+                if let Some(hint) = hint {
+                    write!(f, " — hint: {hint}")?;
+                }
+                Ok(())
+            }
+            Self::ContextWindowExceeded {
+                model,
+                estimated_input_tokens,
+                requested_output_tokens,
+                estimated_total_tokens,
+                context_window_tokens,
+            } => write!(
                f,
-                "missing {provider} credentials; export {} before calling the {provider} API",
-                env_vars.join(" or ")
+                "context_window_blocked for {model}: estimated input {estimated_input_tokens} + requested output {requested_output_tokens} = {estimated_total_tokens} tokens exceeds the {context_window_tokens}-token context window; compact the session or reduce request size before retrying"
            ),
            Self::ExpiredOAuthToken => {
                write!(
@@ -79,19 +281,37 @@ impl Display for ApiError {
            }
            Self::Http(error) => write!(f, "http error: {error}"),
            Self::Io(error) => write!(f, "io error: {error}"),
-            Self::Json(error) => write!(f, "json error: {error}"),
+            Self::Json {
+                provider,
+                model,
+                body_snippet,
+                source,
+            } => write!(
+                f,
+                "failed to parse {provider} response for model {model}: {source}; first 200 chars of body: {body_snippet}"
+            ),
            Self::Api {
                status,
                error_type,
                message,
+                request_id,
                body,
                ..
-            } => match (error_type, message) {
-                (Some(error_type), Some(message)) => {
-                    write!(f, "api returned {status} ({error_type}): {message}")
+            } => {
+                if let (Some(error_type), Some(message)) = (error_type, message) {
+                    write!(f, "api returned {status} ({error_type})")?;
+                    if let Some(request_id) = request_id {
+                        write!(f, " [trace {request_id}]")?;
+                    }
+                    write!(f, ": {message}")
+                } else {
+                    write!(f, "api returned {status}")?;
+                    if let Some(request_id) = request_id {
+                        write!(f, " [trace {request_id}]")?;
+                    }
+                    write!(f, ": {body}")
                }
-                _ => write!(f, "api returned {status}: {body}"),
-            },
+            }
            Self::RetriesExhausted {
                attempts,
                last_error,
@@ -124,7 +344,12 @@ impl From<std::io::Error> for ApiError {

 impl From<serde_json::Error> for ApiError {
    fn from(value: serde_json::Error) -> Self {
-        Self::Json(value)
+        Self::Json {
+            provider: "unknown".to_string(),
+            model: "unknown".to_string(),
+            body_snippet: String::new(),
+            source: value,
+        }
    }
 }

@@ -133,3 +358,215 @@ impl From<VarError> for ApiError {
        Self::InvalidApiKeyEnv(value)
    }
 }
+
+fn looks_like_generic_fatal_wrapper(text: &str) -> bool {
+    let lowered = text.to_ascii_lowercase();
+    GENERIC_FATAL_WRAPPER_MARKERS
+        .iter()
+        .any(|marker| lowered.contains(marker))
+}
+
+fn looks_like_context_window_error(text: &str) -> bool {
+    let lowered = text.to_ascii_lowercase();
+    CONTEXT_WINDOW_ERROR_MARKERS
+        .iter()
+        .any(|marker| lowered.contains(marker))
+}
+
+/// Truncate `body` so the resulting snippet contains at most `max_chars`
+/// characters (counted by Unicode scalar values, not bytes), preserving the
+/// leading slice of the body that the caller most often needs to inspect.
+fn truncate_body_snippet(body: &str, max_chars: usize) -> String {
+    let mut taken_chars = 0;
+    let mut byte_end = 0;
+    for (offset, character) in body.char_indices() {
+        if taken_chars >= max_chars {
+            break;
+        }
+        taken_chars += 1;
+        byte_end = offset + character.len_utf8();
+    }
+    if taken_chars >= max_chars && byte_end < body.len() {
+        format!("{}…", &body[..byte_end])
+    } else {
+        body[..byte_end].to_string()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{truncate_body_snippet, ApiError};
+
+    #[test]
+    fn json_deserialize_error_includes_provider_model_and_truncated_body_snippet() {
+        let raw_body = format!("{}{}", "x".repeat(190), "_TAIL_PAST_200_CHARS_MARKER_");
+        let source = serde_json::from_str::<serde_json::Value>("{not json")
+            .expect_err("invalid json should fail to parse");
+
+        let error = ApiError::json_deserialize("Anthropic", "claude-opus-4-6", &raw_body, source);
+        let rendered = error.to_string();
+
+        assert!(
+            rendered.starts_with("failed to parse Anthropic response for model claude-opus-4-6: "),
+            "rendered error should lead with provider and model: {rendered}"
+        );
+        assert!(
+            rendered.contains("first 200 chars of body: "),
+            "rendered error should label the body snippet: {rendered}"
+        );
+        let snippet = rendered
+            .split("first 200 chars of body: ")
+            .nth(1)
+            .expect("snippet section should be present");
+        assert!(
+            snippet.starts_with(&"x".repeat(190)),
+            "snippet should preserve the leading characters of the body: {snippet}"
+        );
+        assert!(
+            snippet.ends_with('…'),
+            "snippet should signal truncation with an ellipsis: {snippet}"
+        );
+        assert!(
+            !snippet.contains("_TAIL_PAST_200_CHARS_MARKER_"),
+            "snippet should drop characters past the 200-char cap: {snippet}"
+        );
+        assert_eq!(error.safe_failure_class(), "runtime_io");
+        assert_eq!(error.request_id(), None);
+        assert!(!error.is_retryable());
+    }
+
+    #[test]
+    fn truncate_body_snippet_keeps_short_bodies_intact() {
+        assert_eq!(truncate_body_snippet("hello", 200), "hello");
+        assert_eq!(truncate_body_snippet("", 200), "");
+    }
+
+    #[test]
+    fn truncate_body_snippet_caps_long_bodies_at_max_chars() {
+        let body = "a".repeat(250);
+        let snippet = truncate_body_snippet(&body, 200);
+        assert_eq!(snippet.chars().count(), 201, "200 chars + ellipsis");
+        assert!(snippet.ends_with('…'));
+        assert!(snippet.starts_with(&"a".repeat(200)));
+    }
+
+    #[test]
+    fn truncate_body_snippet_does_not_split_multibyte_characters() {
+        let body = "한글한글한글한글한글한글";
+        let snippet = truncate_body_snippet(body, 4);
+        assert_eq!(snippet, "한글한글…");
+    }
+
+    #[test]
+    fn detects_generic_fatal_wrapper_and_classifies_it_as_provider_internal() {
+        let error = ApiError::Api {
+            status: reqwest::StatusCode::INTERNAL_SERVER_ERROR,
+            error_type: Some("api_error".to_string()),
+            message: Some(
+                "Something went wrong while processing your request. Please try again, or use /new to start a fresh session."
+                    .to_string(),
+            ),
+            request_id: Some("req_jobdori_123".to_string()),
+            body: String::new(),
+            retryable: true,
+        };
+
+        assert!(error.is_generic_fatal_wrapper());
+        assert_eq!(error.safe_failure_class(), "provider_internal");
+        assert_eq!(error.request_id(), Some("req_jobdori_123"));
+        assert!(error.to_string().contains("[trace req_jobdori_123]"));
+    }
+
+    #[test]
+    fn retries_exhausted_preserves_nested_request_id_and_failure_class() {
+        let error = ApiError::RetriesExhausted {
+            attempts: 3,
+            last_error: Box::new(ApiError::Api {
+                status: reqwest::StatusCode::BAD_GATEWAY,
+                error_type: Some("api_error".to_string()),
+                message: Some(
+                    "Something went wrong while processing your request. Please try again, or use /new to start a fresh session."
+                        .to_string(),
+                ),
+                request_id: Some("req_nested_456".to_string()),
+                body: String::new(),
+                retryable: true,
+            }),
+        };
+
+        assert!(error.is_generic_fatal_wrapper());
+        assert_eq!(error.safe_failure_class(), "provider_retry_exhausted");
+        assert_eq!(error.request_id(), Some("req_nested_456"));
+    }
+
+    #[test]
+    fn classifies_provider_context_window_errors() {
+        let error = ApiError::Api {
+            status: reqwest::StatusCode::BAD_REQUEST,
+            error_type: Some("invalid_request_error".to_string()),
+            message: Some(
+                "This model's maximum context length is 200000 tokens, but your request used 230000 tokens."
+                    .to_string(),
+            ),
+            request_id: Some("req_ctx_123".to_string()),
+            body: String::new(),
+            retryable: false,
+        };
+
+        assert!(error.is_context_window_failure());
+        assert_eq!(error.safe_failure_class(), "context_window");
+        assert_eq!(error.request_id(), Some("req_ctx_123"));
+    }
+
+    #[test]
+    fn missing_credentials_without_hint_renders_the_canonical_message() {
+        // given
+        let error = ApiError::missing_credentials(
+            "Anthropic",
+            &["ANTHROPIC_AUTH_TOKEN", "ANTHROPIC_API_KEY"],
+        );
+
+        // when
+        let rendered = error.to_string();
+
+        // then
+        assert!(
+            rendered.starts_with(
+                "missing Anthropic credentials; export ANTHROPIC_AUTH_TOKEN or ANTHROPIC_API_KEY before calling the Anthropic API"
+            ),
+            "rendered error should lead with the canonical missing-credential message: {rendered}"
+        );
+        assert!(
+            !rendered.contains(" — hint: "),
+            "no hint should be appended when none is supplied: {rendered}"
+        );
+    }
+
+    #[test]
+    fn missing_credentials_with_hint_appends_the_hint_after_base_message() {
+        // given
+        let error = ApiError::missing_credentials_with_hint(
+            "Anthropic",
+            &["ANTHROPIC_AUTH_TOKEN", "ANTHROPIC_API_KEY"],
+            "I see OPENAI_API_KEY is set — if you meant to use the OpenAI-compat provider, prefix your model name with `openai/` so prefix routing selects it.",
+        );
+
+        // when
+        let rendered = error.to_string();
+
+        // then
+        assert!(
+            rendered.starts_with("missing Anthropic credentials;"),
+            "hint should be appended, not replace the base message: {rendered}"
+        );
+        let hint_marker = " — hint: I see OPENAI_API_KEY is set — if you meant to use the OpenAI-compat provider, prefix your model name with `openai/` so prefix routing selects it.";
+        assert!(
+            rendered.ends_with(hint_marker),
+            "rendered error should end with the hint: {rendered}"
+        );
+        // Classification semantics are unaffected by the presence of a hint.
+        assert_eq!(error.safe_failure_class(), "provider_auth");
+        assert!(!error.is_retryable());
+        assert_eq!(error.request_id(), None);
+    }
+}
@@ -0,0 +1,344 @@
+use crate::error::ApiError;
+
+const HTTP_PROXY_KEYS: [&str; 2] = ["HTTP_PROXY", "http_proxy"];
+const HTTPS_PROXY_KEYS: [&str; 2] = ["HTTPS_PROXY", "https_proxy"];
+const NO_PROXY_KEYS: [&str; 2] = ["NO_PROXY", "no_proxy"];
+
+/// Snapshot of the proxy-related environment variables that influence the
+/// outbound HTTP client. Captured up front so callers can inspect, log, and
+/// test the resolved configuration without re-reading the process environment.
+///
+/// When `proxy_url` is set it acts as a single catch-all proxy for both
+/// HTTP and HTTPS traffic, taking precedence over the per-scheme fields.
+#[derive(Debug, Clone, Default, PartialEq, Eq)]
+pub struct ProxyConfig {
+    pub http_proxy: Option<String>,
+    pub https_proxy: Option<String>,
+    pub no_proxy: Option<String>,
+    /// Optional unified proxy URL that applies to both HTTP and HTTPS.
+    /// When set, this takes precedence over `http_proxy` and `https_proxy`.
+    pub proxy_url: Option<String>,
+}
+
+impl ProxyConfig {
+    /// Read proxy settings from the live process environment, honouring both
+    /// the upper- and lower-case spellings used by curl, git, and friends.
+    #[must_use]
+    pub fn from_env() -> Self {
+        Self::from_lookup(|key| std::env::var(key).ok())
+    }
+
+    /// Create a proxy configuration from a single URL that applies to both
+    /// HTTP and HTTPS traffic. This is the config-file alternative to setting
+    /// `HTTP_PROXY` and `HTTPS_PROXY` environment variables separately.
+    #[must_use]
+    pub fn from_proxy_url(url: impl Into<String>) -> Self {
+        Self {
+            proxy_url: Some(url.into()),
+            ..Self::default()
+        }
+    }
+
+    fn from_lookup<F>(mut lookup: F) -> Self
+    where
+        F: FnMut(&str) -> Option<String>,
+    {
+        Self {
+            http_proxy: first_non_empty(&HTTP_PROXY_KEYS, &mut lookup),
+            https_proxy: first_non_empty(&HTTPS_PROXY_KEYS, &mut lookup),
+            no_proxy: first_non_empty(&NO_PROXY_KEYS, &mut lookup),
+            proxy_url: None,
+        }
+    }
+
+    #[must_use]
+    pub fn is_empty(&self) -> bool {
+        self.proxy_url.is_none() && self.http_proxy.is_none() && self.https_proxy.is_none()
+    }
+}
+
+/// Build a `reqwest::Client` that honours the standard `HTTP_PROXY`,
+/// `HTTPS_PROXY`, and `NO_PROXY` environment variables. When no proxy is
+/// configured the client behaves identically to `reqwest::Client::new()`.
+pub fn build_http_client() -> Result<reqwest::Client, ApiError> {
+    build_http_client_with(&ProxyConfig::from_env())
+}
+
+/// Infallible counterpart to [`build_http_client`] for constructors that
+/// historically returned `Self` rather than `Result<Self, _>`. When the proxy
+/// configuration is malformed we fall back to a default client so that
+/// callers retain the previous behaviour and the failure surfaces on the
+/// first outbound request instead of at construction time.
+#[must_use]
+pub fn build_http_client_or_default() -> reqwest::Client {
+    build_http_client().unwrap_or_else(|_| reqwest::Client::new())
+}
+
+/// Build a `reqwest::Client` from an explicit [`ProxyConfig`]. Used by tests
+/// and by callers that want to override process-level environment lookups.
+///
+/// When `config.proxy_url` is set it overrides the per-scheme `http_proxy`
+/// and `https_proxy` fields and is registered as both an HTTP and HTTPS
+/// proxy so a single value can route every outbound request.
+pub fn build_http_client_with(config: &ProxyConfig) -> Result<reqwest::Client, ApiError> {
+    let mut builder = reqwest::Client::builder().no_proxy();
+
+    let no_proxy = config
+        .no_proxy
+        .as_deref()
+        .and_then(reqwest::NoProxy::from_string);
+
+    let (http_proxy_url, https_proxy_url) = match config.proxy_url.as_deref() {
+        Some(unified) => (Some(unified), Some(unified)),
+        None => (config.http_proxy.as_deref(), config.https_proxy.as_deref()),
+    };
+
+    if let Some(url) = https_proxy_url {
+        let mut proxy = reqwest::Proxy::https(url)?;
+        if let Some(filter) = no_proxy.clone() {
+            proxy = proxy.no_proxy(Some(filter));
+        }
+        builder = builder.proxy(proxy);
+    }
+
+    if let Some(url) = http_proxy_url {
+        let mut proxy = reqwest::Proxy::http(url)?;
+        if let Some(filter) = no_proxy.clone() {
+            proxy = proxy.no_proxy(Some(filter));
+        }
+        builder = builder.proxy(proxy);
+    }
+
+    Ok(builder.build()?)
+}
+
+fn first_non_empty<F>(keys: &[&str], lookup: &mut F) -> Option<String>
+where
+    F: FnMut(&str) -> Option<String>,
+{
+    keys.iter()
+        .find_map(|key| lookup(key).filter(|value| !value.is_empty()))
+}
+
+#[cfg(test)]
+mod tests {
+    use std::collections::HashMap;
+
+    use super::{build_http_client_with, ProxyConfig};
+
+    fn config_from_map(pairs: &[(&str, &str)]) -> ProxyConfig {
+        let map: HashMap<String, String> = pairs
+            .iter()
+            .map(|(key, value)| ((*key).to_string(), (*value).to_string()))
+            .collect();
+        ProxyConfig::from_lookup(|key| map.get(key).cloned())
+    }
+
+    #[test]
+    fn proxy_config_is_empty_when_no_env_vars_are_set() {
+        // given
+        let config = config_from_map(&[]);
+
+        // when
+        let empty = config.is_empty();
+
+        // then
+        assert!(empty);
+        assert_eq!(config, ProxyConfig::default());
+    }
+
+    #[test]
+    fn proxy_config_reads_uppercase_http_https_and_no_proxy() {
+        // given
+        let pairs = [
+            ("HTTP_PROXY", "http://proxy.internal:3128"),
+            ("HTTPS_PROXY", "http://secure.internal:3129"),
+            ("NO_PROXY", "localhost,127.0.0.1,.corp"),
+        ];
+
+        // when
+        let config = config_from_map(&pairs);
+
+        // then
+        assert_eq!(
+            config.http_proxy.as_deref(),
+            Some("http://proxy.internal:3128")
+        );
+        assert_eq!(
+            config.https_proxy.as_deref(),
+            Some("http://secure.internal:3129")
+        );
+        assert_eq!(
+            config.no_proxy.as_deref(),
+            Some("localhost,127.0.0.1,.corp")
+        );
+        assert!(!config.is_empty());
+    }
+
+    #[test]
+    fn proxy_config_falls_back_to_lowercase_keys() {
+        // given
+        let pairs = [
+            ("http_proxy", "http://lower.internal:3128"),
+            ("https_proxy", "http://lower-secure.internal:3129"),
+            ("no_proxy", ".lower"),
+        ];
+
+        // when
+        let config = config_from_map(&pairs);
+
+        // then
+        assert_eq!(
+            config.http_proxy.as_deref(),
+            Some("http://lower.internal:3128")
+        );
+        assert_eq!(
+            config.https_proxy.as_deref(),
+            Some("http://lower-secure.internal:3129")
+        );
+        assert_eq!(config.no_proxy.as_deref(), Some(".lower"));
+    }
+
+    #[test]
+    fn proxy_config_prefers_uppercase_over_lowercase_when_both_set() {
+        // given
+        let pairs = [
+            ("HTTP_PROXY", "http://upper.internal:3128"),
+            ("http_proxy", "http://lower.internal:3128"),
+        ];
+
+        // when
+        let config = config_from_map(&pairs);
+
+        // then
+        assert_eq!(
+            config.http_proxy.as_deref(),
+            Some("http://upper.internal:3128")
+        );
+    }
+
+    #[test]
+    fn proxy_config_treats_empty_strings_as_unset() {
+        // given
+        let pairs = [("HTTP_PROXY", ""), ("http_proxy", "")];
+
+        // when
+        let config = config_from_map(&pairs);
+
+        // then
+        assert!(config.http_proxy.is_none());
+    }
+
+    #[test]
+    fn build_http_client_succeeds_when_no_proxy_is_configured() {
+        // given
+        let config = ProxyConfig::default();
+
+        // when
+        let result = build_http_client_with(&config);
+
+        // then
+        assert!(result.is_ok());
+    }
+
+    #[test]
+    fn build_http_client_succeeds_with_valid_http_and_https_proxies() {
+        // given
+        let config = ProxyConfig {
+            http_proxy: Some("http://proxy.internal:3128".to_string()),
+            https_proxy: Some("http://secure.internal:3129".to_string()),
+            no_proxy: Some("localhost,127.0.0.1".to_string()),
+            proxy_url: None,
+        };
+
+        // when
+        let result = build_http_client_with(&config);
+
+        // then
+        assert!(result.is_ok());
+    }
+
+    #[test]
+    fn build_http_client_returns_http_error_for_invalid_proxy_url() {
+        // given
+        let config = ProxyConfig {
+            http_proxy: None,
+            https_proxy: Some("not a url".to_string()),
+            no_proxy: None,
+            proxy_url: None,
+        };
+
+        // when
+        let result = build_http_client_with(&config);
+
+        // then
+        let error = result.expect_err("invalid proxy URL must be reported as a build failure");
+        assert!(
+            matches!(error, crate::error::ApiError::Http(_)),
+            "expected ApiError::Http for invalid proxy URL, got: {error:?}"
+        );
+    }
+
+    #[test]
+    fn from_proxy_url_sets_unified_field_and_leaves_per_scheme_empty() {
+        // given / when
+        let config = ProxyConfig::from_proxy_url("http://unified.internal:3128");
+
+        // then
+        assert_eq!(
+            config.proxy_url.as_deref(),
+            Some("http://unified.internal:3128")
+        );
+        assert!(config.http_proxy.is_none());
+        assert!(config.https_proxy.is_none());
+        assert!(!config.is_empty());
+    }
+
+    #[test]
+    fn build_http_client_succeeds_with_unified_proxy_url() {
+        // given
+        let config = ProxyConfig {
+            proxy_url: Some("http://unified.internal:3128".to_string()),
+            no_proxy: Some("localhost".to_string()),
+            ..ProxyConfig::default()
+        };
+
+        // when
+        let result = build_http_client_with(&config);
+
+        // then
+        assert!(result.is_ok());
+    }
+
+    #[test]
+    fn proxy_url_takes_precedence_over_per_scheme_fields() {
+        // given – both per-scheme and unified are set
+        let config = ProxyConfig {
+            http_proxy: Some("http://per-scheme.internal:1111".to_string()),
+            https_proxy: Some("http://per-scheme.internal:2222".to_string()),
+            no_proxy: None,
+            proxy_url: Some("http://unified.internal:3128".to_string()),
+        };
+
+        // when – building succeeds (the unified URL is valid)
+        let result = build_http_client_with(&config);
+
+        // then
+        assert!(result.is_ok());
+    }
+
+    #[test]
+    fn build_http_client_returns_error_for_invalid_unified_proxy_url() {
+        // given
+        let config = ProxyConfig::from_proxy_url("not a url");
+
+        // when
+        let result = build_http_client_with(&config);
+
+        // then
+        assert!(
+            matches!(result, Err(crate::error::ApiError::Http(_))),
+            "invalid unified proxy URL should fail: {result:?}"
+        );
+    }
+}
@@ -1,5 +1,7 @@
 mod client;
 mod error;
+mod http_client;
+mod prompt_cache;
 mod providers;
 mod sse;
 mod types;
@@ -9,10 +11,18 @@ pub use client::{
    resolve_startup_auth_source, MessageStream, OAuthTokenSet, ProviderClient,
 };
 pub use error::ApiError;
-pub use providers::claw_provider::{AuthSource, ClawApiClient, ClawApiClient as ApiClient};
+pub use http_client::{
+    build_http_client, build_http_client_or_default, build_http_client_with, ProxyConfig,
+};
+pub use prompt_cache::{
+    CacheBreakEvent, PromptCache, PromptCacheConfig, PromptCachePaths, PromptCacheRecord,
+    PromptCacheStats,
+};
+pub use providers::anthropic::{AnthropicClient, AnthropicClient as ApiClient, AuthSource};
 pub use providers::openai_compat::{OpenAiCompatClient, OpenAiCompatConfig};
 pub use providers::{
-    detect_provider_kind, max_tokens_for_model, resolve_model_alias, ProviderKind,
+    detect_provider_kind, max_tokens_for_model, max_tokens_for_model_with_override,
+    resolve_model_alias, ProviderKind,
 };
 pub use sse::{parse_frame, SseParser};
 pub use types::{
@@ -21,3 +31,9 @@ pub use types::{
    MessageResponse, MessageStartEvent, MessageStopEvent, OutputContentBlock, StreamEvent,
    ToolChoice, ToolDefinition, ToolResultContentBlock, Usage,
 };
+
+pub use telemetry::{
+    AnalyticsEvent, AnthropicRequestProfile, ClientIdentity, JsonlTelemetrySink,
+    MemoryTelemetrySink, SessionTraceRecord, SessionTracer, TelemetryEvent, TelemetrySink,
+    DEFAULT_ANTHROPIC_VERSION,
+};
@@ -0,0 +1,735 @@
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::sync::{Arc, Mutex};
+use std::time::{Duration, SystemTime, UNIX_EPOCH};
+
+use serde::{Deserialize, Serialize};
+
+use crate::types::{MessageRequest, MessageResponse, Usage};
+
+const DEFAULT_COMPLETION_TTL_SECS: u64 = 30;
+const DEFAULT_PROMPT_TTL_SECS: u64 = 5 * 60;
+const DEFAULT_BREAK_MIN_DROP: u32 = 2_000;
+const MAX_SANITIZED_LENGTH: usize = 80;
+const REQUEST_FINGERPRINT_VERSION: u32 = 1;
+const REQUEST_FINGERPRINT_PREFIX: &str = "v1";
+const FNV_OFFSET_BASIS: u64 = 0xcbf2_9ce4_8422_2325;
+const FNV_PRIME: u64 = 0x0000_0100_0000_01b3;
+
+#[derive(Debug, Clone)]
+pub struct PromptCacheConfig {
+    pub session_id: String,
+    pub completion_ttl: Duration,
+    pub prompt_ttl: Duration,
+    pub cache_break_min_drop: u32,
+}
+
+impl PromptCacheConfig {
+    #[must_use]
+    pub fn new(session_id: impl Into<String>) -> Self {
+        Self {
+            session_id: session_id.into(),
+            completion_ttl: Duration::from_secs(DEFAULT_COMPLETION_TTL_SECS),
+            prompt_ttl: Duration::from_secs(DEFAULT_PROMPT_TTL_SECS),
+            cache_break_min_drop: DEFAULT_BREAK_MIN_DROP,
+        }
+    }
+}
+
+impl Default for PromptCacheConfig {
+    fn default() -> Self {
+        Self::new("default")
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct PromptCachePaths {
+    pub root: PathBuf,
+    pub session_dir: PathBuf,
+    pub completion_dir: PathBuf,
+    pub session_state_path: PathBuf,
+    pub stats_path: PathBuf,
+}
+
+impl PromptCachePaths {
+    #[must_use]
+    pub fn for_session(session_id: &str) -> Self {
+        let root = base_cache_root();
+        let session_dir = root.join(sanitize_path_segment(session_id));
+        let completion_dir = session_dir.join("completions");
+        Self {
+            root,
+            session_state_path: session_dir.join("session-state.json"),
+            stats_path: session_dir.join("stats.json"),
+            session_dir,
+            completion_dir,
+        }
+    }
+
+    #[must_use]
+    pub fn completion_entry_path(&self, request_hash: &str) -> PathBuf {
+        self.completion_dir.join(format!("{request_hash}.json"))
+    }
+}
+
+#[derive(Debug, Clone, Default, PartialEq, Eq, Serialize, Deserialize)]
+pub struct PromptCacheStats {
+    pub tracked_requests: u64,
+    pub completion_cache_hits: u64,
+    pub completion_cache_misses: u64,
+    pub completion_cache_writes: u64,
+    pub expected_invalidations: u64,
+    pub unexpected_cache_breaks: u64,
+    pub total_cache_creation_input_tokens: u64,
+    pub total_cache_read_input_tokens: u64,
+    pub last_cache_creation_input_tokens: Option<u32>,
+    pub last_cache_read_input_tokens: Option<u32>,
+    pub last_request_hash: Option<String>,
+    pub last_completion_cache_key: Option<String>,
+    pub last_break_reason: Option<String>,
+    pub last_cache_source: Option<String>,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct CacheBreakEvent {
+    pub unexpected: bool,
+    pub reason: String,
+    pub previous_cache_read_input_tokens: u32,
+    pub current_cache_read_input_tokens: u32,
+    pub token_drop: u32,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct PromptCacheRecord {
+    pub cache_break: Option<CacheBreakEvent>,
+    pub stats: PromptCacheStats,
+}
+
+#[derive(Debug, Clone)]
+pub struct PromptCache {
+    inner: Arc<Mutex<PromptCacheInner>>,
+}
+
+impl PromptCache {
+    #[must_use]
+    pub fn new(session_id: impl Into<String>) -> Self {
+        Self::with_config(PromptCacheConfig::new(session_id))
+    }
+
+    #[must_use]
+    pub fn with_config(config: PromptCacheConfig) -> Self {
+        let paths = PromptCachePaths::for_session(&config.session_id);
+        let stats = read_json::<PromptCacheStats>(&paths.stats_path).unwrap_or_default();
+        let previous = read_json::<TrackedPromptState>(&paths.session_state_path);
+        Self {
+            inner: Arc::new(Mutex::new(PromptCacheInner {
+                config,
+                paths,
+                stats,
+                previous,
+            })),
+        }
+    }
+
+    #[must_use]
+    pub fn paths(&self) -> PromptCachePaths {
+        self.lock().paths.clone()
+    }
+
+    #[must_use]
+    pub fn stats(&self) -> PromptCacheStats {
+        self.lock().stats.clone()
+    }
+
+    #[must_use]
+    pub fn lookup_completion(&self, request: &MessageRequest) -> Option<MessageResponse> {
+        let request_hash = request_hash_hex(request);
+        let (paths, ttl) = {
+            let inner = self.lock();
+            (inner.paths.clone(), inner.config.completion_ttl)
+        };
+        let entry_path = paths.completion_entry_path(&request_hash);
+        let entry = read_json::<CompletionCacheEntry>(&entry_path);
+        let Some(entry) = entry else {
+            let mut inner = self.lock();
+            inner.stats.completion_cache_misses += 1;
+            inner.stats.last_completion_cache_key = Some(request_hash);
+            persist_state(&inner);
+            return None;
+        };
+
+        if entry.fingerprint_version != current_fingerprint_version() {
+            let mut inner = self.lock();
+            inner.stats.completion_cache_misses += 1;
+            inner.stats.last_completion_cache_key = Some(request_hash.clone());
+            let _ = fs::remove_file(entry_path);
+            persist_state(&inner);
+            return None;
+        }
+
+        let expired = now_unix_secs().saturating_sub(entry.cached_at_unix_secs) >= ttl.as_secs();
+        let mut inner = self.lock();
+        inner.stats.last_completion_cache_key = Some(request_hash.clone());
+        if expired {
+            inner.stats.completion_cache_misses += 1;
+            let _ = fs::remove_file(entry_path);
+            persist_state(&inner);
+            return None;
+        }
+
+        inner.stats.completion_cache_hits += 1;
+        apply_usage_to_stats(
+            &mut inner.stats,
+            &entry.response.usage,
+            &request_hash,
+            "completion-cache",
+        );
+        inner.previous = Some(TrackedPromptState::from_usage(
+            request,
+            &entry.response.usage,
+        ));
+        persist_state(&inner);
+        Some(entry.response)
+    }
+
+    #[must_use]
+    pub fn record_response(
+        &self,
+        request: &MessageRequest,
+        response: &MessageResponse,
+    ) -> PromptCacheRecord {
+        self.record_usage_internal(request, &response.usage, Some(response))
+    }
+
+    #[must_use]
+    pub fn record_usage(&self, request: &MessageRequest, usage: &Usage) -> PromptCacheRecord {
+        self.record_usage_internal(request, usage, None)
+    }
+
+    fn record_usage_internal(
+        &self,
+        request: &MessageRequest,
+        usage: &Usage,
+        response: Option<&MessageResponse>,
+    ) -> PromptCacheRecord {
+        let request_hash = request_hash_hex(request);
+        let mut inner = self.lock();
+        let previous = inner.previous.clone();
+        let current = TrackedPromptState::from_usage(request, usage);
+        let cache_break = detect_cache_break(&inner.config, previous.as_ref(), &current);
+
+        inner.stats.tracked_requests += 1;
+        apply_usage_to_stats(&mut inner.stats, usage, &request_hash, "api-response");
+        if let Some(event) = &cache_break {
+            if event.unexpected {
+                inner.stats.unexpected_cache_breaks += 1;
+            } else {
+                inner.stats.expected_invalidations += 1;
+            }
+            inner.stats.last_break_reason = Some(event.reason.clone());
+        }
+
+        inner.previous = Some(current);
+        if let Some(response) = response {
+            write_completion_entry(&inner.paths, &request_hash, response);
+            inner.stats.completion_cache_writes += 1;
+        }
+        persist_state(&inner);
+
+        PromptCacheRecord {
+            cache_break,
+            stats: inner.stats.clone(),
+        }
+    }
+
+    fn lock(&self) -> std::sync::MutexGuard<'_, PromptCacheInner> {
+        self.inner
+            .lock()
+            .unwrap_or_else(std::sync::PoisonError::into_inner)
+    }
+}
+
+#[derive(Debug)]
+struct PromptCacheInner {
+    config: PromptCacheConfig,
+    paths: PromptCachePaths,
+    stats: PromptCacheStats,
+    previous: Option<TrackedPromptState>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+struct CompletionCacheEntry {
+    cached_at_unix_secs: u64,
+    #[serde(default = "current_fingerprint_version")]
+    fingerprint_version: u32,
+    response: MessageResponse,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+struct TrackedPromptState {
+    observed_at_unix_secs: u64,
+    #[serde(default = "current_fingerprint_version")]
+    fingerprint_version: u32,
+    model_hash: u64,
+    system_hash: u64,
+    tools_hash: u64,
+    messages_hash: u64,
+    cache_read_input_tokens: u32,
+}
+
+impl TrackedPromptState {
+    fn from_usage(request: &MessageRequest, usage: &Usage) -> Self {
+        let hashes = RequestFingerprints::from_request(request);
+        Self {
+            observed_at_unix_secs: now_unix_secs(),
+            fingerprint_version: current_fingerprint_version(),
+            model_hash: hashes.model,
+            system_hash: hashes.system,
+            tools_hash: hashes.tools,
+            messages_hash: hashes.messages,
+            cache_read_input_tokens: usage.cache_read_input_tokens,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy)]
+struct RequestFingerprints {
+    model: u64,
+    system: u64,
+    tools: u64,
+    messages: u64,
+}
+
+impl RequestFingerprints {
+    fn from_request(request: &MessageRequest) -> Self {
+        Self {
+            model: hash_serializable(&request.model),
+            system: hash_serializable(&request.system),
+            tools: hash_serializable(&request.tools),
+            messages: hash_serializable(&request.messages),
+        }
+    }
+}
+
+fn detect_cache_break(
+    config: &PromptCacheConfig,
+    previous: Option<&TrackedPromptState>,
+    current: &TrackedPromptState,
+) -> Option<CacheBreakEvent> {
+    let previous = previous?;
+    if previous.fingerprint_version != current.fingerprint_version {
+        return Some(CacheBreakEvent {
+            unexpected: false,
+            reason: format!(
+                "fingerprint version changed (v{} -> v{})",
+                previous.fingerprint_version, current.fingerprint_version
+            ),
+            previous_cache_read_input_tokens: previous.cache_read_input_tokens,
+            current_cache_read_input_tokens: current.cache_read_input_tokens,
+            token_drop: previous
+                .cache_read_input_tokens
+                .saturating_sub(current.cache_read_input_tokens),
+        });
+    }
+    let token_drop = previous
+        .cache_read_input_tokens
+        .saturating_sub(current.cache_read_input_tokens);
+    if token_drop < config.cache_break_min_drop {
+        return None;
+    }
+
+    let mut reasons = Vec::new();
+    if previous.model_hash != current.model_hash {
+        reasons.push("model changed");
+    }
+    if previous.system_hash != current.system_hash {
+        reasons.push("system prompt changed");
+    }
+    if previous.tools_hash != current.tools_hash {
+        reasons.push("tool definitions changed");
+    }
+    if previous.messages_hash != current.messages_hash {
+        reasons.push("message payload changed");
+    }
+
+    let elapsed = current
+        .observed_at_unix_secs
+        .saturating_sub(previous.observed_at_unix_secs);
+
+    let (unexpected, reason) = if reasons.is_empty() {
+        if elapsed > config.prompt_ttl.as_secs() {
+            (
+                false,
+                format!("possible prompt cache TTL expiry after {elapsed}s"),
+            )
+        } else {
+            (
+                true,
+                "cache read tokens dropped while prompt fingerprint remained stable".to_string(),
+            )
+        }
+    } else {
+        (false, reasons.join(", "))
+    };
+
+    Some(CacheBreakEvent {
+        unexpected,
+        reason,
+        previous_cache_read_input_tokens: previous.cache_read_input_tokens,
+        current_cache_read_input_tokens: current.cache_read_input_tokens,
+        token_drop,
+    })
+}
+
+fn apply_usage_to_stats(
+    stats: &mut PromptCacheStats,
+    usage: &Usage,
+    request_hash: &str,
+    source: &str,
+) {
+    stats.total_cache_creation_input_tokens += u64::from(usage.cache_creation_input_tokens);
+    stats.total_cache_read_input_tokens += u64::from(usage.cache_read_input_tokens);
+    stats.last_cache_creation_input_tokens = Some(usage.cache_creation_input_tokens);
+    stats.last_cache_read_input_tokens = Some(usage.cache_read_input_tokens);
+    stats.last_request_hash = Some(request_hash.to_string());
+    stats.last_cache_source = Some(source.to_string());
+}
+
+fn persist_state(inner: &PromptCacheInner) {
+    let _ = ensure_cache_dirs(&inner.paths);
+    let _ = write_json(&inner.paths.stats_path, &inner.stats);
+    if let Some(previous) = &inner.previous {
+        let _ = write_json(&inner.paths.session_state_path, previous);
+    }
+}
+
+fn write_completion_entry(
+    paths: &PromptCachePaths,
+    request_hash: &str,
+    response: &MessageResponse,
+) {
+    let _ = ensure_cache_dirs(paths);
+    let entry = CompletionCacheEntry {
+        cached_at_unix_secs: now_unix_secs(),
+        fingerprint_version: current_fingerprint_version(),
+        response: response.clone(),
+    };
+    let _ = write_json(&paths.completion_entry_path(request_hash), &entry);
+}
+
+fn ensure_cache_dirs(paths: &PromptCachePaths) -> std::io::Result<()> {
+    fs::create_dir_all(&paths.completion_dir)
+}
+
+fn write_json<T: Serialize>(path: &Path, value: &T) -> std::io::Result<()> {
+    let json = serde_json::to_vec_pretty(value)
+        .map_err(|error| std::io::Error::new(std::io::ErrorKind::InvalidData, error))?;
+    fs::write(path, json)
+}
+
+fn read_json<T: for<'de> Deserialize<'de>>(path: &Path) -> Option<T> {
+    let bytes = fs::read(path).ok()?;
+    serde_json::from_slice(&bytes).ok()
+}
+
+fn request_hash_hex(request: &MessageRequest) -> String {
+    format!(
+        "{REQUEST_FINGERPRINT_PREFIX}-{:016x}",
+        hash_serializable(request)
+    )
+}
+
+fn hash_serializable<T: Serialize>(value: &T) -> u64 {
+    let json = serde_json::to_vec(value).unwrap_or_default();
+    stable_hash_bytes(&json)
+}
+
+fn sanitize_path_segment(value: &str) -> String {
+    let sanitized: String = value
+        .chars()
+        .map(|ch| if ch.is_ascii_alphanumeric() { ch } else { '-' })
+        .collect();
+    if sanitized.len() <= MAX_SANITIZED_LENGTH {
+        return sanitized;
+    }
+    let suffix = format!("-{:x}", hash_string(value));
+    format!(
+        "{}{}",
+        &sanitized[..MAX_SANITIZED_LENGTH.saturating_sub(suffix.len())],
+        suffix
+    )
+}
+
+fn hash_string(value: &str) -> u64 {
+    stable_hash_bytes(value.as_bytes())
+}
+
+fn base_cache_root() -> PathBuf {
+    if let Some(config_home) = std::env::var_os("CLAUDE_CONFIG_HOME") {
+        return PathBuf::from(config_home)
+            .join("cache")
+            .join("prompt-cache");
+    }
+    if let Some(home) = std::env::var_os("HOME") {
+        return PathBuf::from(home)
+            .join(".claude")
+            .join("cache")
+            .join("prompt-cache");
+    }
+    std::env::temp_dir().join("claude-prompt-cache")
+}
+
+fn now_unix_secs() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .map_or(0, |duration| duration.as_secs())
+}
+
+const fn current_fingerprint_version() -> u32 {
+    REQUEST_FINGERPRINT_VERSION
+}
+
+fn stable_hash_bytes(bytes: &[u8]) -> u64 {
+    let mut hash = FNV_OFFSET_BASIS;
+    for byte in bytes {
+        hash ^= u64::from(*byte);
+        hash = hash.wrapping_mul(FNV_PRIME);
+    }
+    hash
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::{Mutex, OnceLock};
+    use std::time::{Duration, SystemTime, UNIX_EPOCH};
+
+    use super::{
+        detect_cache_break, read_json, request_hash_hex, sanitize_path_segment, PromptCache,
+        PromptCacheConfig, PromptCachePaths, TrackedPromptState, REQUEST_FINGERPRINT_PREFIX,
+    };
+    use crate::types::{InputMessage, MessageRequest, MessageResponse, OutputContentBlock, Usage};
+
+    fn test_env_lock() -> std::sync::MutexGuard<'static, ()> {
+        static LOCK: OnceLock<Mutex<()>> = OnceLock::new();
+        LOCK.get_or_init(|| Mutex::new(()))
+            .lock()
+            .unwrap_or_else(std::sync::PoisonError::into_inner)
+    }
+
+    #[test]
+    fn path_builder_sanitizes_session_identifier() {
+        let paths = PromptCachePaths::for_session("session:/with spaces");
+        let session_dir = paths
+            .session_dir
+            .file_name()
+            .and_then(|value| value.to_str())
+            .expect("session dir name");
+        assert_eq!(session_dir, "session--with-spaces");
+        assert!(paths.completion_dir.ends_with("completions"));
+        assert!(paths.stats_path.ends_with("stats.json"));
+        assert!(paths.session_state_path.ends_with("session-state.json"));
+    }
+
+    #[test]
+    fn request_fingerprint_drives_unexpected_break_detection() {
+        let request = sample_request("same");
+        let previous = TrackedPromptState::from_usage(
+            &request,
+            &Usage {
+                input_tokens: 0,
+                cache_creation_input_tokens: 0,
+                cache_read_input_tokens: 6_000,
+                output_tokens: 0,
+            },
+        );
+        let current = TrackedPromptState::from_usage(
+            &request,
+            &Usage {
+                input_tokens: 0,
+                cache_creation_input_tokens: 0,
+                cache_read_input_tokens: 1_000,
+                output_tokens: 0,
+            },
+        );
+        let event = detect_cache_break(&PromptCacheConfig::default(), Some(&previous), &current)
+            .expect("break should be detected");
+        assert!(event.unexpected);
+        assert!(event.reason.contains("stable"));
+    }
+
+    #[test]
+    fn changed_prompt_marks_break_as_expected() {
+        let previous_request = sample_request("first");
+        let current_request = sample_request("second");
+        let previous = TrackedPromptState::from_usage(
+            &previous_request,
+            &Usage {
+                input_tokens: 0,
+                cache_creation_input_tokens: 0,
+                cache_read_input_tokens: 6_000,
+                output_tokens: 0,
+            },
+        );
+        let current = TrackedPromptState::from_usage(
+            &current_request,
+            &Usage {
+                input_tokens: 0,
+                cache_creation_input_tokens: 0,
+                cache_read_input_tokens: 1_000,
+                output_tokens: 0,
+            },
+        );
+        let event = detect_cache_break(&PromptCacheConfig::default(), Some(&previous), &current)
+            .expect("break should be detected");
+        assert!(!event.unexpected);
+        assert!(event.reason.contains("message payload changed"));
+    }
+
+    #[test]
+    fn completion_cache_round_trip_persists_recent_response() {
+        let _guard = test_env_lock();
+        let temp_root = std::env::temp_dir().join(format!(
+            "prompt-cache-test-{}-{}",
+            std::process::id(),
+            SystemTime::now()
+                .duration_since(UNIX_EPOCH)
+                .expect("time")
+                .as_nanos()
+        ));
+        std::env::set_var("CLAUDE_CONFIG_HOME", &temp_root);
+        let cache = PromptCache::new("unit-test-session");
+        let request = sample_request("cache me");
+        let response = sample_response(42, 12, "cached");
+
+        assert!(cache.lookup_completion(&request).is_none());
+        let record = cache.record_response(&request, &response);
+        assert!(record.cache_break.is_none());
+
+        let cached = cache
+            .lookup_completion(&request)
+            .expect("cached response should load");
+        assert_eq!(cached.content, response.content);
+
+        let stats = cache.stats();
+        assert_eq!(stats.completion_cache_hits, 1);
+        assert_eq!(stats.completion_cache_misses, 1);
+        assert_eq!(stats.completion_cache_writes, 1);
+
+        let persisted = read_json::<super::PromptCacheStats>(&cache.paths().stats_path)
+            .expect("stats should persist");
+        assert_eq!(persisted.completion_cache_hits, 1);
+
+        std::fs::remove_dir_all(temp_root).expect("cleanup temp root");
+        std::env::remove_var("CLAUDE_CONFIG_HOME");
+    }
+
+    #[test]
+    fn distinct_requests_do_not_collide_in_completion_cache() {
+        let _guard = test_env_lock();
+        let temp_root = std::env::temp_dir().join(format!(
+            "prompt-cache-distinct-{}-{}",
+            std::process::id(),
+            SystemTime::now()
+                .duration_since(UNIX_EPOCH)
+                .expect("time")
+                .as_nanos()
+        ));
+        std::env::set_var("CLAUDE_CONFIG_HOME", &temp_root);
+        let cache = PromptCache::new("distinct-request-session");
+        let first_request = sample_request("first");
+        let second_request = sample_request("second");
+
+        let response = sample_response(42, 12, "cached");
+        let _ = cache.record_response(&first_request, &response);
+
+        assert!(cache.lookup_completion(&second_request).is_none());
+
+        std::fs::remove_dir_all(temp_root).expect("cleanup temp root");
+        std::env::remove_var("CLAUDE_CONFIG_HOME");
+    }
+
+    #[test]
+    fn expired_completion_entries_are_not_reused() {
+        let _guard = test_env_lock();
+        let temp_root = std::env::temp_dir().join(format!(
+            "prompt-cache-expired-{}-{}",
+            std::process::id(),
+            SystemTime::now()
+                .duration_since(UNIX_EPOCH)
+                .expect("time")
+                .as_nanos()
+        ));
+        std::env::set_var("CLAUDE_CONFIG_HOME", &temp_root);
+        let cache = PromptCache::with_config(PromptCacheConfig {
+            session_id: "expired-session".to_string(),
+            completion_ttl: Duration::ZERO,
+            ..PromptCacheConfig::default()
+        });
+        let request = sample_request("expire me");
+        let response = sample_response(7, 3, "stale");
+
+        let _ = cache.record_response(&request, &response);
+
+        assert!(cache.lookup_completion(&request).is_none());
+        let stats = cache.stats();
+        assert_eq!(stats.completion_cache_hits, 0);
+        assert_eq!(stats.completion_cache_misses, 1);
+
+        std::fs::remove_dir_all(temp_root).expect("cleanup temp root");
+        std::env::remove_var("CLAUDE_CONFIG_HOME");
+    }
+
+    #[test]
+    fn sanitize_path_caps_long_values() {
+        let long_value = "x".repeat(200);
+        let sanitized = sanitize_path_segment(&long_value);
+        assert!(sanitized.len() <= 80);
+    }
+
+    #[test]
+    fn request_hashes_are_versioned_and_stable() {
+        let request = sample_request("stable");
+        let first = request_hash_hex(&request);
+        let second = request_hash_hex(&request);
+        assert_eq!(first, second);
+        assert!(first.starts_with(REQUEST_FINGERPRINT_PREFIX));
+    }
+
+    fn sample_request(text: &str) -> MessageRequest {
+        MessageRequest {
+            model: "claude-3-7-sonnet-latest".to_string(),
+            max_tokens: 64,
+            messages: vec![InputMessage::user_text(text)],
+            system: Some("system".to_string()),
+            tools: None,
+            tool_choice: None,
+            stream: false,
+            ..Default::default()
+        }
+    }
+
+    fn sample_response(
+        cache_read_input_tokens: u32,
+        output_tokens: u32,
+        text: &str,
+    ) -> MessageResponse {
+        MessageResponse {
+            id: "msg_test".to_string(),
+            kind: "message".to_string(),
+            role: "assistant".to_string(),
+            content: vec![OutputContentBlock::Text {
+                text: text.to_string(),
+            }],
+            model: "claude-3-7-sonnet-latest".to_string(),
+            stop_reason: Some("end_turn".to_string()),
+            stop_sequence: None,
+            usage: Usage {
+                input_tokens: 10,
+                cache_creation_input_tokens: 5,
+                cache_read_input_tokens,
+                output_tokens,
+            },
+            request_id: Some("req_test".to_string()),
+        }
+    }
+}
@@ -1,14 +1,19 @@
+#![allow(clippy::cast_possible_truncation)]
 use std::future::Future;
 use std::pin::Pin;

+use serde::Serialize;
+
 use crate::error::ApiError;
 use crate::types::{MessageRequest, MessageResponse};

-pub mod claw_provider;
+pub mod anthropic;
 pub mod openai_compat;

+#[allow(dead_code)]
 pub type ProviderFuture<'a, T> = Pin<Box<dyn Future<Output = Result<T, ApiError>> + Send + 'a>>;

+#[allow(dead_code)]
 pub trait Provider {
    type Stream;

@@ -25,7 +30,7 @@ pub trait Provider {

 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub enum ProviderKind {
-    ClawApi,
+    Anthropic,
    Xai,
    OpenAi,
 }
@@ -38,59 +43,38 @@ pub struct ProviderMetadata {
    pub default_base_url: &'static str,
 }

+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub struct ModelTokenLimit {
+    pub max_output_tokens: u32,
+    pub context_window_tokens: u32,
+}
+
 const MODEL_REGISTRY: &[(&str, ProviderMetadata)] = &[
    (
        "opus",
        ProviderMetadata {
-            provider: ProviderKind::ClawApi,
+            provider: ProviderKind::Anthropic,
            auth_env: "ANTHROPIC_API_KEY",
            base_url_env: "ANTHROPIC_BASE_URL",
-            default_base_url: claw_provider::DEFAULT_BASE_URL,
+            default_base_url: anthropic::DEFAULT_BASE_URL,
        },
    ),
    (
        "sonnet",
        ProviderMetadata {
-            provider: ProviderKind::ClawApi,
+            provider: ProviderKind::Anthropic,
            auth_env: "ANTHROPIC_API_KEY",
            base_url_env: "ANTHROPIC_BASE_URL",
-            default_base_url: claw_provider::DEFAULT_BASE_URL,
+            default_base_url: anthropic::DEFAULT_BASE_URL,
        },
    ),
    (
        "haiku",
        ProviderMetadata {
-            provider: ProviderKind::ClawApi,
+            provider: ProviderKind::Anthropic,
            auth_env: "ANTHROPIC_API_KEY",
            base_url_env: "ANTHROPIC_BASE_URL",
-            default_base_url: claw_provider::DEFAULT_BASE_URL,
-        },
-    ),
-    (
-        "claude-opus-4-6",
-        ProviderMetadata {
-            provider: ProviderKind::ClawApi,
-            auth_env: "ANTHROPIC_API_KEY",
-            base_url_env: "ANTHROPIC_BASE_URL",
-            default_base_url: claw_provider::DEFAULT_BASE_URL,
-        },
-    ),
-    (
-        "claude-sonnet-4-6",
-        ProviderMetadata {
-            provider: ProviderKind::ClawApi,
-            auth_env: "ANTHROPIC_API_KEY",
-            base_url_env: "ANTHROPIC_BASE_URL",
-            default_base_url: claw_provider::DEFAULT_BASE_URL,
-        },
-    ),
-    (
-        "claude-haiku-4-5-20251213",
-        ProviderMetadata {
-            provider: ProviderKind::ClawApi,
-            auth_env: "ANTHROPIC_API_KEY",
-            base_url_env: "ANTHROPIC_BASE_URL",
-            default_base_url: claw_provider::DEFAULT_BASE_URL,
+            default_base_url: anthropic::DEFAULT_BASE_URL,
        },
    ),
    (
@@ -148,7 +132,7 @@ pub fn resolve_model_alias(model: &str) -> String {
        .iter()
        .find_map(|(alias, metadata)| {
            (*alias == lower).then_some(match metadata.provider {
-                ProviderKind::ClawApi => match *alias {
+                ProviderKind::Anthropic => match *alias {
                    "opus" => "claude-opus-4-6",
                    "sonnet" => "claude-sonnet-4-6",
                    "haiku" => "claude-haiku-4-5-20251213",
@@ -169,11 +153,15 @@ pub fn resolve_model_alias(model: &str) -> String {
 #[must_use]
 pub fn metadata_for_model(model: &str) -> Option<ProviderMetadata> {
    let canonical = resolve_model_alias(model);
-    let lower = canonical.to_ascii_lowercase();
-    if let Some((_, metadata)) = MODEL_REGISTRY.iter().find(|(alias, _)| *alias == lower) {
-        return Some(*metadata);
+    if canonical.starts_with("claude") {
+        return Some(ProviderMetadata {
+            provider: ProviderKind::Anthropic,
+            auth_env: "ANTHROPIC_API_KEY",
+            base_url_env: "ANTHROPIC_BASE_URL",
+            default_base_url: anthropic::DEFAULT_BASE_URL,
+        });
    }
-    if lower.starts_with("grok") {
+    if canonical.starts_with("grok") {
        return Some(ProviderMetadata {
            provider: ProviderKind::Xai,
            auth_env: "XAI_API_KEY",
@@ -181,6 +169,31 @@ pub fn metadata_for_model(model: &str) -> Option<ProviderMetadata> {
            default_base_url: openai_compat::DEFAULT_XAI_BASE_URL,
        });
    }
+    // Explicit provider-namespaced models (e.g. "openai/gpt-4.1-mini") must
+    // route to the correct provider regardless of which auth env vars are set.
+    // Without this, detect_provider_kind falls through to the auth-sniffer
+    // order and misroutes to Anthropic if ANTHROPIC_API_KEY is present.
+    if canonical.starts_with("openai/") || canonical.starts_with("gpt-") {
+        return Some(ProviderMetadata {
+            provider: ProviderKind::OpenAi,
+            auth_env: "OPENAI_API_KEY",
+            base_url_env: "OPENAI_BASE_URL",
+            default_base_url: openai_compat::DEFAULT_OPENAI_BASE_URL,
+        });
+    }
+    // Alibaba DashScope compatible-mode endpoint. Routes qwen/* and bare
+    // qwen-* model names (qwen-max, qwen-plus, qwen-turbo, qwen-qwq, etc.)
+    // to the OpenAI-compat client pointed at DashScope's /compatible-mode/v1.
+    // Uses the OpenAi provider kind because DashScope speaks the OpenAI REST
+    // shape — only the base URL and auth env var differ.
+    if canonical.starts_with("qwen/") || canonical.starts_with("qwen-") {
+        return Some(ProviderMetadata {
+            provider: ProviderKind::OpenAi,
+            auth_env: "DASHSCOPE_API_KEY",
+            base_url_env: "DASHSCOPE_BASE_URL",
+            default_base_url: openai_compat::DEFAULT_DASHSCOPE_BASE_URL,
+        });
+    }
    None
 }

@@ -189,8 +202,17 @@ pub fn detect_provider_kind(model: &str) -> ProviderKind {
    if let Some(metadata) = metadata_for_model(model) {
        return metadata.provider;
    }
-    if claw_provider::has_auth_from_env_or_saved().unwrap_or(false) {
-        return ProviderKind::ClawApi;
+    // When OPENAI_BASE_URL is set, the user explicitly configured an
+    // OpenAI-compatible endpoint. Prefer it over the Anthropic fallback
+    // even when the model name has no recognized prefix — this is the
+    // common case for local providers (Ollama, LM Studio, vLLM, etc.)
+    // where model names like "qwen2.5-coder:7b" don't match any prefix.
+    if std::env::var_os("OPENAI_BASE_URL").is_some() && openai_compat::has_api_key("OPENAI_API_KEY")
+    {
+        return ProviderKind::OpenAi;
+    }
+    if anthropic::has_auth_from_env_or_saved().unwrap_or(false) {
+        return ProviderKind::Anthropic;
    }
    if openai_compat::has_api_key("OPENAI_API_KEY") {
        return ProviderKind::OpenAi;
@@ -198,22 +220,271 @@ pub fn detect_provider_kind(model: &str) -> ProviderKind {
    if openai_compat::has_api_key("XAI_API_KEY") {
        return ProviderKind::Xai;
    }
-    ProviderKind::ClawApi
+    // Last resort: if OPENAI_BASE_URL is set without OPENAI_API_KEY (some
+    // local providers like Ollama don't require auth), still route there.
+    if std::env::var_os("OPENAI_BASE_URL").is_some() {
+        return ProviderKind::OpenAi;
+    }
+    ProviderKind::Anthropic
 }

 #[must_use]
 pub fn max_tokens_for_model(model: &str) -> u32 {
+    model_token_limit(model).map_or_else(
+        || {
+            let canonical = resolve_model_alias(model);
+            if canonical.contains("opus") {
+                32_000
+            } else {
+                64_000
+            }
+        },
+        |limit| limit.max_output_tokens,
+    )
+}
+
+/// Returns the effective max output tokens for a model, preferring a plugin
+/// override when present. Falls back to [`max_tokens_for_model`] when the
+/// override is `None`.
+#[must_use]
+pub fn max_tokens_for_model_with_override(model: &str, plugin_override: Option<u32>) -> u32 {
+    plugin_override.unwrap_or_else(|| max_tokens_for_model(model))
+}
+
+#[must_use]
+pub fn model_token_limit(model: &str) -> Option<ModelTokenLimit> {
    let canonical = resolve_model_alias(model);
-    if canonical.contains("opus") {
-        32_000
-    } else {
-        64_000
+    match canonical.as_str() {
+        "claude-opus-4-6" => Some(ModelTokenLimit {
+            max_output_tokens: 32_000,
+            context_window_tokens: 200_000,
+        }),
+        "claude-sonnet-4-6" | "claude-haiku-4-5-20251213" => Some(ModelTokenLimit {
+            max_output_tokens: 64_000,
+            context_window_tokens: 200_000,
+        }),
+        "grok-3" | "grok-3-mini" => Some(ModelTokenLimit {
+            max_output_tokens: 64_000,
+            context_window_tokens: 131_072,
+        }),
+        _ => None,
    }
 }

+pub fn preflight_message_request(request: &MessageRequest) -> Result<(), ApiError> {
+    let Some(limit) = model_token_limit(&request.model) else {
+        return Ok(());
+    };
+
+    let estimated_input_tokens = estimate_message_request_input_tokens(request);
+    let estimated_total_tokens = estimated_input_tokens.saturating_add(request.max_tokens);
+    if estimated_total_tokens > limit.context_window_tokens {
+        return Err(ApiError::ContextWindowExceeded {
+            model: resolve_model_alias(&request.model),
+            estimated_input_tokens,
+            requested_output_tokens: request.max_tokens,
+            estimated_total_tokens,
+            context_window_tokens: limit.context_window_tokens,
+        });
+    }
+
+    Ok(())
+}
+
+fn estimate_message_request_input_tokens(request: &MessageRequest) -> u32 {
+    let mut estimate = estimate_serialized_tokens(&request.messages);
+    estimate = estimate.saturating_add(estimate_serialized_tokens(&request.system));
+    estimate = estimate.saturating_add(estimate_serialized_tokens(&request.tools));
+    estimate = estimate.saturating_add(estimate_serialized_tokens(&request.tool_choice));
+    estimate
+}
+
+fn estimate_serialized_tokens<T: Serialize>(value: &T) -> u32 {
+    serde_json::to_vec(value)
+        .ok()
+        .map_or(0, |bytes| (bytes.len() / 4 + 1) as u32)
+}
+
+/// Env var names used by other provider backends. When Anthropic auth
+/// resolution fails we sniff these so we can hint the user that their
+/// credentials probably belong to a different provider and suggest the
+/// model-prefix routing fix that would select it.
+const FOREIGN_PROVIDER_ENV_VARS: &[(&str, &str, &str)] = &[
+    (
+        "OPENAI_API_KEY",
+        "OpenAI-compat",
+        "prefix your model name with `openai/` (e.g. `--model openai/gpt-4.1-mini`) so prefix routing selects the OpenAI-compatible provider, and set `OPENAI_BASE_URL` if you are pointing at OpenRouter/Ollama/a local server",
+    ),
+    (
+        "XAI_API_KEY",
+        "xAI",
+        "use an xAI model alias (e.g. `--model grok` or `--model grok-mini`) so the prefix router selects the xAI backend",
+    ),
+    (
+        "DASHSCOPE_API_KEY",
+        "Alibaba DashScope",
+        "prefix your model name with `qwen/` or `qwen-` (e.g. `--model qwen-plus`) so prefix routing selects the DashScope backend",
+    ),
+];
+
+/// Check whether an env var is set to a non-empty value either in the real
+/// process environment or in the working-directory `.env` file. Mirrors the
+/// credential discovery path used by `read_env_non_empty` so the hint text
+/// stays truthful when users rely on `.env` instead of a real export.
+fn env_or_dotenv_present(key: &str) -> bool {
+    match std::env::var(key) {
+        Ok(value) if !value.is_empty() => true,
+        Ok(_) | Err(std::env::VarError::NotPresent) => {
+            dotenv_value(key).is_some_and(|value| !value.is_empty())
+        }
+        Err(_) => false,
+    }
+}
+
+/// Produce a hint string describing the first foreign provider credential
+/// that is present in the environment when Anthropic auth resolution has
+/// just failed. Returns `None` when no foreign credential is set, in which
+/// case the caller should fall back to the plain `missing_credentials`
+/// error without a hint.
+pub(crate) fn anthropic_missing_credentials_hint() -> Option<String> {
+    for (env_var, provider_label, fix_hint) in FOREIGN_PROVIDER_ENV_VARS {
+        if env_or_dotenv_present(env_var) {
+            return Some(format!(
+                "I see {env_var} is set — if you meant to use the {provider_label} provider, {fix_hint}."
+            ));
+        }
+    }
+    None
+}
+
+/// Build an Anthropic-specific `MissingCredentials` error, attaching a
+/// hint suggesting the probable fix whenever a different provider's
+/// credentials are already present in the environment. Anthropic call
+/// sites should prefer this helper over `ApiError::missing_credentials`
+/// so users who mistyped a model name or forgot the prefix get a useful
+/// signal instead of a generic "missing Anthropic credentials" wall.
+pub(crate) fn anthropic_missing_credentials() -> ApiError {
+    const PROVIDER: &str = "Anthropic";
+    const ENV_VARS: &[&str] = &["ANTHROPIC_AUTH_TOKEN", "ANTHROPIC_API_KEY"];
+    match anthropic_missing_credentials_hint() {
+        Some(hint) => ApiError::missing_credentials_with_hint(PROVIDER, ENV_VARS, hint),
+        None => ApiError::missing_credentials(PROVIDER, ENV_VARS),
+    }
+}
+
+/// Parse a `.env` file body into key/value pairs using a minimal `KEY=VALUE`
+/// grammar. Lines that are blank, start with `#`, or do not contain `=` are
+/// ignored. Surrounding double or single quotes are stripped from the value.
+/// An optional leading `export ` prefix on the key is also stripped so files
+/// shared with shell `source` workflows still parse cleanly.
+pub(crate) fn parse_dotenv(content: &str) -> std::collections::HashMap<String, String> {
+    let mut values = std::collections::HashMap::new();
+    for raw_line in content.lines() {
+        let line = raw_line.trim();
+        if line.is_empty() || line.starts_with('#') {
+            continue;
+        }
+        let Some((raw_key, raw_value)) = line.split_once('=') else {
+            continue;
+        };
+        let trimmed_key = raw_key.trim();
+        let key = trimmed_key
+            .strip_prefix("export ")
+            .map_or(trimmed_key, str::trim)
+            .to_string();
+        if key.is_empty() {
+            continue;
+        }
+        let trimmed_value = raw_value.trim();
+        let unquoted = if (trimmed_value.starts_with('"') && trimmed_value.ends_with('"')
+            || trimmed_value.starts_with('\'') && trimmed_value.ends_with('\''))
+            && trimmed_value.len() >= 2
+        {
+            &trimmed_value[1..trimmed_value.len() - 1]
+        } else {
+            trimmed_value
+        };
+        values.insert(key, unquoted.to_string());
+    }
+    values
+}
+
+/// Load and parse a `.env` file from the given path. Missing files yield
+/// `None` instead of an error so callers can use this as a soft fallback.
+pub(crate) fn load_dotenv_file(
+    path: &std::path::Path,
+) -> Option<std::collections::HashMap<String, String>> {
+    let content = std::fs::read_to_string(path).ok()?;
+    Some(parse_dotenv(&content))
+}
+
+/// Look up `key` in a `.env` file located in the current working directory.
+/// Returns `None` when the file is missing, the key is absent, or the value
+/// is empty.
+pub(crate) fn dotenv_value(key: &str) -> Option<String> {
+    let cwd = std::env::current_dir().ok()?;
+    let values = load_dotenv_file(&cwd.join(".env"))?;
+    values.get(key).filter(|value| !value.is_empty()).cloned()
+}
+
 #[cfg(test)]
 mod tests {
-    use super::{detect_provider_kind, max_tokens_for_model, resolve_model_alias, ProviderKind};
+    use std::ffi::OsString;
+    use std::sync::{Mutex, OnceLock};
+
+    use serde_json::json;
+
+    use crate::error::ApiError;
+    use crate::types::{
+        InputContentBlock, InputMessage, MessageRequest, ToolChoice, ToolDefinition,
+    };
+
+    use super::{
+        anthropic_missing_credentials, anthropic_missing_credentials_hint, detect_provider_kind,
+        load_dotenv_file, max_tokens_for_model, max_tokens_for_model_with_override,
+        model_token_limit, parse_dotenv, preflight_message_request, resolve_model_alias,
+        ProviderKind,
+    };
+
+    /// Serializes every test in this module that mutates process-wide
+    /// environment variables so concurrent test threads cannot observe
+    /// each other's partially-applied state while probing the foreign
+    /// provider credential sniffer.
+    fn env_lock() -> std::sync::MutexGuard<'static, ()> {
+        static LOCK: OnceLock<Mutex<()>> = OnceLock::new();
+        LOCK.get_or_init(|| Mutex::new(()))
+            .lock()
+            .unwrap_or_else(std::sync::PoisonError::into_inner)
+    }
+
+    /// Snapshot-restore guard for a single environment variable. Captures
+    /// the original value on construction, applies the requested override
+    /// (set or remove), and restores the original on drop so tests leave
+    /// the process env untouched even when they panic mid-assertion.
+    struct EnvVarGuard {
+        key: &'static str,
+        original: Option<OsString>,
+    }
+
+    impl EnvVarGuard {
+        fn set(key: &'static str, value: Option<&str>) -> Self {
+            let original = std::env::var_os(key);
+            match value {
+                Some(value) => std::env::set_var(key, value),
+                None => std::env::remove_var(key),
+            }
+            Self { key, original }
+        }
+    }
+
+    impl Drop for EnvVarGuard {
+        fn drop(&mut self) {
+            match self.original.take() {
+                Some(value) => std::env::set_var(self.key, value),
+                None => std::env::remove_var(self.key),
+            }
+        }
+    }

    #[test]
    fn resolves_grok_aliases() {
@@ -227,7 +498,59 @@ mod tests {
        assert_eq!(detect_provider_kind("grok"), ProviderKind::Xai);
        assert_eq!(
            detect_provider_kind("claude-sonnet-4-6"),
-            ProviderKind::ClawApi
+            ProviderKind::Anthropic
+        );
+    }
+
+    #[test]
+    fn openai_namespaced_model_routes_to_openai_not_anthropic() {
+        // Regression: "openai/gpt-4.1-mini" was misrouted to Anthropic when
+        // ANTHROPIC_API_KEY was set because metadata_for_model returned None
+        // and detect_provider_kind fell through to auth-sniffer order.
+        // The model prefix must win over env-var presence.
+        let kind = super::metadata_for_model("openai/gpt-4.1-mini")
+            .map(|m| m.provider)
+            .unwrap_or_else(|| detect_provider_kind("openai/gpt-4.1-mini"));
+        assert_eq!(
+            kind,
+            ProviderKind::OpenAi,
+            "openai/ prefix must route to OpenAi regardless of ANTHROPIC_API_KEY"
+        );
+
+        // Also cover bare gpt- prefix
+        let kind2 = super::metadata_for_model("gpt-4o")
+            .map(|m| m.provider)
+            .unwrap_or_else(|| detect_provider_kind("gpt-4o"));
+        assert_eq!(kind2, ProviderKind::OpenAi);
+    }
+
+    #[test]
+    fn qwen_prefix_routes_to_dashscope_not_anthropic() {
+        // User request from Discord #clawcode-get-help: web3g wants to use
+        // Qwen 3.6 Plus via native Alibaba DashScope API (not OpenRouter,
+        // which has lower rate limits). metadata_for_model must route
+        // qwen/* and bare qwen-* to the OpenAi provider kind pointed at
+        // the DashScope compatible-mode endpoint, regardless of whether
+        // ANTHROPIC_API_KEY is present in the environment.
+        let meta = super::metadata_for_model("qwen/qwen-max")
+            .expect("qwen/ prefix must resolve to DashScope metadata");
+        assert_eq!(meta.provider, ProviderKind::OpenAi);
+        assert_eq!(meta.auth_env, "DASHSCOPE_API_KEY");
+        assert_eq!(meta.base_url_env, "DASHSCOPE_BASE_URL");
+        assert!(meta.default_base_url.contains("dashscope.aliyuncs.com"));
+
+        // Bare qwen- prefix also routes
+        let meta2 = super::metadata_for_model("qwen-plus")
+            .expect("qwen- prefix must resolve to DashScope metadata");
+        assert_eq!(meta2.provider, ProviderKind::OpenAi);
+        assert_eq!(meta2.auth_env, "DASHSCOPE_API_KEY");
+
+        // detect_provider_kind must agree even if ANTHROPIC_API_KEY is set
+        let kind = detect_provider_kind("qwen/qwen3-coder");
+        assert_eq!(
+            kind,
+            ProviderKind::OpenAi,
+            "qwen/ prefix must win over auth-sniffer order"
        );
    }

@@ -236,4 +559,467 @@ mod tests {
        assert_eq!(max_tokens_for_model("opus"), 32_000);
        assert_eq!(max_tokens_for_model("grok-3"), 64_000);
    }
+
+    #[test]
+    fn plugin_config_max_output_tokens_overrides_model_default() {
+        // given
+        let nanos = std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_nanos();
+        let root = std::env::temp_dir().join(format!("api-plugin-max-tokens-{nanos}"));
+        let cwd = root.join("project");
+        let home = root.join("home").join(".claw");
+        std::fs::create_dir_all(cwd.join(".claw")).expect("project config dir");
+        std::fs::create_dir_all(&home).expect("home config dir");
+        std::fs::write(
+            home.join("settings.json"),
+            r#"{
+              "plugins": {
+                "maxOutputTokens": 12345
+              }
+            }"#,
+        )
+        .expect("write plugin settings");
+
+        // when
+        let loaded = runtime::ConfigLoader::new(&cwd, &home)
+            .load()
+            .expect("config should load");
+        let plugin_override = loaded.plugins().max_output_tokens();
+        let effective = max_tokens_for_model_with_override("claude-opus-4-6", plugin_override);
+
+        // then
+        assert_eq!(plugin_override, Some(12345));
+        assert_eq!(effective, 12345);
+        assert_ne!(effective, max_tokens_for_model("claude-opus-4-6"));
+
+        std::fs::remove_dir_all(root).expect("cleanup temp dir");
+    }
+
+    #[test]
+    fn max_tokens_for_model_with_override_falls_back_when_plugin_unset() {
+        // given
+        let plugin_override: Option<u32> = None;
+
+        // when
+        let effective = max_tokens_for_model_with_override("claude-opus-4-6", plugin_override);
+
+        // then
+        assert_eq!(effective, max_tokens_for_model("claude-opus-4-6"));
+        assert_eq!(effective, 32_000);
+    }
+
+    #[test]
+    fn returns_context_window_metadata_for_supported_models() {
+        assert_eq!(
+            model_token_limit("claude-sonnet-4-6")
+                .expect("claude-sonnet-4-6 should be registered")
+                .context_window_tokens,
+            200_000
+        );
+        assert_eq!(
+            model_token_limit("grok-mini")
+                .expect("grok-mini should resolve to a registered model")
+                .context_window_tokens,
+            131_072
+        );
+    }
+
+    #[test]
+    fn preflight_blocks_requests_that_exceed_the_model_context_window() {
+        let request = MessageRequest {
+            model: "claude-sonnet-4-6".to_string(),
+            max_tokens: 64_000,
+            messages: vec![InputMessage {
+                role: "user".to_string(),
+                content: vec![InputContentBlock::Text {
+                    text: "x".repeat(600_000),
+                }],
+            }],
+            system: Some("Keep the answer short.".to_string()),
+            tools: Some(vec![ToolDefinition {
+                name: "weather".to_string(),
+                description: Some("Fetches weather".to_string()),
+                input_schema: json!({
+                    "type": "object",
+                    "properties": { "city": { "type": "string" } },
+                }),
+            }]),
+            tool_choice: Some(ToolChoice::Auto),
+            stream: true,
+            ..Default::default()
+        };
+
+        let error = preflight_message_request(&request)
+            .expect_err("oversized request should be rejected before the provider call");
+
+        match error {
+            ApiError::ContextWindowExceeded {
+                model,
+                estimated_input_tokens,
+                requested_output_tokens,
+                estimated_total_tokens,
+                context_window_tokens,
+            } => {
+                assert_eq!(model, "claude-sonnet-4-6");
+                assert!(estimated_input_tokens > 136_000);
+                assert_eq!(requested_output_tokens, 64_000);
+                assert!(estimated_total_tokens > context_window_tokens);
+                assert_eq!(context_window_tokens, 200_000);
+            }
+            other => panic!("expected context-window preflight failure, got {other:?}"),
+        }
+    }
+
+    #[test]
+    fn preflight_skips_unknown_models() {
+        let request = MessageRequest {
+            model: "unknown-model".to_string(),
+            max_tokens: 64_000,
+            messages: vec![InputMessage {
+                role: "user".to_string(),
+                content: vec![InputContentBlock::Text {
+                    text: "x".repeat(600_000),
+                }],
+            }],
+            system: None,
+            tools: None,
+            tool_choice: None,
+            stream: false,
+            ..Default::default()
+        };
+
+        preflight_message_request(&request)
+            .expect("models without context metadata should skip the guarded preflight");
+    }
+
+    #[test]
+    fn parse_dotenv_extracts_keys_handles_comments_quotes_and_export_prefix() {
+        // given
+        let body = "\
+# this is a comment
+
+ANTHROPIC_API_KEY=plain-value
+XAI_API_KEY=\"quoted-value\"
+OPENAI_API_KEY='single-quoted'
+export GROK_API_KEY=exported-value
+   PADDED_KEY  =  padded-value  
+EMPTY_VALUE=
+NO_EQUALS_LINE
+";
+
+        // when
+        let values = parse_dotenv(body);
+
+        // then
+        assert_eq!(
+            values.get("ANTHROPIC_API_KEY").map(String::as_str),
+            Some("plain-value")
+        );
+        assert_eq!(
+            values.get("XAI_API_KEY").map(String::as_str),
+            Some("quoted-value")
+        );
+        assert_eq!(
+            values.get("OPENAI_API_KEY").map(String::as_str),
+            Some("single-quoted")
+        );
+        assert_eq!(
+            values.get("GROK_API_KEY").map(String::as_str),
+            Some("exported-value")
+        );
+        assert_eq!(
+            values.get("PADDED_KEY").map(String::as_str),
+            Some("padded-value")
+        );
+        assert_eq!(values.get("EMPTY_VALUE").map(String::as_str), Some(""));
+        assert!(!values.contains_key("NO_EQUALS_LINE"));
+        assert!(!values.contains_key("# this is a comment"));
+    }
+
+    #[test]
+    fn load_dotenv_file_reads_keys_from_disk_and_returns_none_when_missing() {
+        // given
+        let temp_root = std::env::temp_dir().join(format!(
+            "api-dotenv-test-{}-{}",
+            std::process::id(),
+            std::time::SystemTime::now()
+                .duration_since(std::time::UNIX_EPOCH)
+                .map_or(0, |duration| duration.as_nanos())
+        ));
+        std::fs::create_dir_all(&temp_root).expect("create temp dir");
+        let env_path = temp_root.join(".env");
+        std::fs::write(
+            &env_path,
+            "ANTHROPIC_API_KEY=secret-from-file\n# comment\nXAI_API_KEY=\"xai-secret\"\n",
+        )
+        .expect("write .env");
+        let missing_path = temp_root.join("does-not-exist.env");
+
+        // when
+        let loaded = load_dotenv_file(&env_path).expect("file should load");
+        let missing = load_dotenv_file(&missing_path);
+
+        // then
+        assert_eq!(
+            loaded.get("ANTHROPIC_API_KEY").map(String::as_str),
+            Some("secret-from-file")
+        );
+        assert_eq!(
+            loaded.get("XAI_API_KEY").map(String::as_str),
+            Some("xai-secret")
+        );
+        assert!(missing.is_none());
+
+        let _ = std::fs::remove_dir_all(&temp_root);
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_hint_is_none_when_no_foreign_creds_present() {
+        // given
+        let _lock = env_lock();
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", None);
+        let _xai = EnvVarGuard::set("XAI_API_KEY", None);
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", None);
+
+        // when
+        let hint = anthropic_missing_credentials_hint();
+
+        // then
+        assert!(
+            hint.is_none(),
+            "no hint should be produced when every foreign provider env var is absent, got {hint:?}"
+        );
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_hint_detects_openai_api_key_and_recommends_openai_prefix() {
+        // given
+        let _lock = env_lock();
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", Some("sk-openrouter-varleg"));
+        let _xai = EnvVarGuard::set("XAI_API_KEY", None);
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", None);
+
+        // when
+        let hint = anthropic_missing_credentials_hint()
+            .expect("OPENAI_API_KEY presence should produce a hint");
+
+        // then
+        assert!(
+            hint.contains("OPENAI_API_KEY is set"),
+            "hint should name the detected env var so users recognize it: {hint}"
+        );
+        assert!(
+            hint.contains("OpenAI-compat"),
+            "hint should identify the target provider: {hint}"
+        );
+        assert!(
+            hint.contains("openai/"),
+            "hint should mention the `openai/` prefix routing fix: {hint}"
+        );
+        assert!(
+            hint.contains("OPENAI_BASE_URL"),
+            "hint should mention OPENAI_BASE_URL so OpenRouter users see the full picture: {hint}"
+        );
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_hint_detects_xai_api_key() {
+        // given
+        let _lock = env_lock();
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", None);
+        let _xai = EnvVarGuard::set("XAI_API_KEY", Some("xai-test-key"));
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", None);
+
+        // when
+        let hint = anthropic_missing_credentials_hint()
+            .expect("XAI_API_KEY presence should produce a hint");
+
+        // then
+        assert!(
+            hint.contains("XAI_API_KEY is set"),
+            "hint should name XAI_API_KEY: {hint}"
+        );
+        assert!(
+            hint.contains("xAI"),
+            "hint should identify the xAI provider: {hint}"
+        );
+        assert!(
+            hint.contains("grok"),
+            "hint should suggest a grok-prefixed model alias: {hint}"
+        );
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_hint_detects_dashscope_api_key() {
+        // given
+        let _lock = env_lock();
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", None);
+        let _xai = EnvVarGuard::set("XAI_API_KEY", None);
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", Some("sk-dashscope-test"));
+
+        // when
+        let hint = anthropic_missing_credentials_hint()
+            .expect("DASHSCOPE_API_KEY presence should produce a hint");
+
+        // then
+        assert!(
+            hint.contains("DASHSCOPE_API_KEY is set"),
+            "hint should name DASHSCOPE_API_KEY: {hint}"
+        );
+        assert!(
+            hint.contains("DashScope"),
+            "hint should identify the DashScope provider: {hint}"
+        );
+        assert!(
+            hint.contains("qwen"),
+            "hint should suggest a qwen-prefixed model alias: {hint}"
+        );
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_hint_prefers_openai_when_multiple_foreign_creds_set() {
+        // given
+        let _lock = env_lock();
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", Some("sk-openrouter-varleg"));
+        let _xai = EnvVarGuard::set("XAI_API_KEY", Some("xai-test-key"));
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", Some("sk-dashscope-test"));
+
+        // when
+        let hint = anthropic_missing_credentials_hint()
+            .expect("multiple foreign creds should still produce a hint");
+
+        // then
+        assert!(
+            hint.contains("OPENAI_API_KEY"),
+            "OpenAI should be prioritized because it is the most common misrouting pattern (OpenRouter users), got: {hint}"
+        );
+        assert!(
+            !hint.contains("XAI_API_KEY"),
+            "only the first detected provider should be named to keep the hint focused, got: {hint}"
+        );
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_builds_error_with_canonical_env_vars_and_no_hint_when_clean() {
+        // given
+        let _lock = env_lock();
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", None);
+        let _xai = EnvVarGuard::set("XAI_API_KEY", None);
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", None);
+
+        // when
+        let error = anthropic_missing_credentials();
+
+        // then
+        match &error {
+            ApiError::MissingCredentials {
+                provider,
+                env_vars,
+                hint,
+            } => {
+                assert_eq!(*provider, "Anthropic");
+                assert_eq!(*env_vars, &["ANTHROPIC_AUTH_TOKEN", "ANTHROPIC_API_KEY"]);
+                assert!(
+                    hint.is_none(),
+                    "clean environment should not generate a hint, got {hint:?}"
+                );
+            }
+            other => panic!("expected MissingCredentials variant, got {other:?}"),
+        }
+        let rendered = error.to_string();
+        assert!(
+            !rendered.contains(" — hint: "),
+            "rendered error should be a plain missing-creds message: {rendered}"
+        );
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_builds_error_with_hint_when_openai_key_is_set() {
+        // given
+        let _lock = env_lock();
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", Some("sk-openrouter-varleg"));
+        let _xai = EnvVarGuard::set("XAI_API_KEY", None);
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", None);
+
+        // when
+        let error = anthropic_missing_credentials();
+
+        // then
+        match &error {
+            ApiError::MissingCredentials {
+                provider,
+                env_vars,
+                hint,
+            } => {
+                assert_eq!(*provider, "Anthropic");
+                assert_eq!(*env_vars, &["ANTHROPIC_AUTH_TOKEN", "ANTHROPIC_API_KEY"]);
+                let hint_value = hint.as_deref().expect("hint should be populated");
+                assert!(
+                    hint_value.contains("OPENAI_API_KEY is set"),
+                    "hint should name the detected env var: {hint_value}"
+                );
+            }
+            other => panic!("expected MissingCredentials variant, got {other:?}"),
+        }
+        let rendered = error.to_string();
+        assert!(
+            rendered.starts_with("missing Anthropic credentials;"),
+            "canonical base message should still lead the rendered error: {rendered}"
+        );
+        assert!(
+            rendered.contains(" — hint: I see OPENAI_API_KEY is set"),
+            "rendered error should carry the env-driven hint: {rendered}"
+        );
+    }
+
+    #[test]
+    fn anthropic_missing_credentials_hint_ignores_empty_string_values() {
+        // given
+        let _lock = env_lock();
+        // An empty value is semantically equivalent to "not set" for the
+        // credential discovery path, so the sniffer must treat it that way
+        // to avoid false-positive hints for users who intentionally cleared
+        // a stale export with `OPENAI_API_KEY=`.
+        let _openai = EnvVarGuard::set("OPENAI_API_KEY", Some(""));
+        let _xai = EnvVarGuard::set("XAI_API_KEY", None);
+        let _dashscope = EnvVarGuard::set("DASHSCOPE_API_KEY", None);
+
+        // when
+        let hint = anthropic_missing_credentials_hint();
+
+        // then
+        assert!(
+            hint.is_none(),
+            "empty env var should not trigger the hint sniffer, got {hint:?}"
+        );
+    }
+
+    #[test]
+    fn openai_base_url_overrides_anthropic_fallback_for_unknown_model() {
+        // given — user has OPENAI_BASE_URL + OPENAI_API_KEY but no Anthropic
+        // creds, and a model name with no recognized prefix.
+        let _lock = env_lock();
+        let _base_url = EnvVarGuard::set("OPENAI_BASE_URL", Some("http://127.0.0.1:11434/v1"));
+        let _api_key = EnvVarGuard::set("OPENAI_API_KEY", Some("dummy"));
+        let _anthropic_key = EnvVarGuard::set("ANTHROPIC_API_KEY", None);
+        let _anthropic_token = EnvVarGuard::set("ANTHROPIC_AUTH_TOKEN", None);
+
+        // when
+        let provider = detect_provider_kind("qwen2.5-coder:7b");
+
+        // then — should route to OpenAI, not Anthropic
+        assert_eq!(
+            provider,
+            ProviderKind::OpenAi,
+            "OPENAI_BASE_URL should win over Anthropic fallback for unknown models"
+        );
+    }
+
+    // NOTE: a "OPENAI_BASE_URL without OPENAI_API_KEY" test is omitted
+    // because workspace-parallel test binaries can race on process env
+    // (env_lock only protects within a single binary). The detection logic
+    // is covered: OPENAI_BASE_URL alone routes to OpenAi as a last-resort
+    // fallback in detect_provider_kind().
 }
@@ -4,6 +4,8 @@ use crate::types::StreamEvent;
 #[derive(Debug, Default)]
 pub struct SseParser {
    buffer: Vec<u8>,
+    provider: Option<String>,
+    model: Option<String>,
 }

 impl SseParser {
@@ -12,12 +14,23 @@ impl SseParser {
        Self::default()
    }

+    /// Attach the provider name and model to this parser so that JSON
+    /// deserialization failures within streamed frames carry enough context
+    /// for callers to understand which upstream produced the unparseable
+    /// payload.
+    #[must_use]
+    pub fn with_context(mut self, provider: impl Into<String>, model: impl Into<String>) -> Self {
+        self.provider = Some(provider.into());
+        self.model = Some(model.into());
+        self
+    }
+
    pub fn push(&mut self, chunk: &[u8]) -> Result<Vec<StreamEvent>, ApiError> {
        self.buffer.extend_from_slice(chunk);
        let mut events = Vec::new();

        while let Some(frame) = self.next_frame() {
-            if let Some(event) = parse_frame(&frame)? {
+            if let Some(event) = self.parse_frame_with_context(&frame)? {
                events.push(event);
            }
        }
@@ -31,12 +44,18 @@ impl SseParser {
        }

        let trailing = std::mem::take(&mut self.buffer);
-        match parse_frame(&String::from_utf8_lossy(&trailing))? {
+        match self.parse_frame_with_context(&String::from_utf8_lossy(&trailing))? {
            Some(event) => Ok(vec![event]),
            None => Ok(Vec::new()),
        }
    }

+    fn parse_frame_with_context(&self, frame: &str) -> Result<Option<StreamEvent>, ApiError> {
+        let provider = self.provider.as_deref().unwrap_or("unknown");
+        let model = self.model.as_deref().unwrap_or("unknown");
+        parse_frame_with_provider(frame, provider, model)
+    }
+
    fn next_frame(&mut self) -> Option<String> {
        let separator = self
            .buffer
@@ -61,6 +80,14 @@ impl SseParser {
 }

 pub fn parse_frame(frame: &str) -> Result<Option<StreamEvent>, ApiError> {
+    parse_frame_with_provider(frame, "unknown", "unknown")
+}
+
+pub(crate) fn parse_frame_with_provider(
+    frame: &str,
+    provider: &str,
+    model: &str,
+) -> Result<Option<StreamEvent>, ApiError> {
    let trimmed = frame.trim();
    if trimmed.is_empty() {
        return Ok(None);
@@ -97,7 +124,7 @@ pub fn parse_frame(frame: &str) -> Result<Option<StreamEvent>, ApiError> {

    serde_json::from_str::<StreamEvent>(&payload)
        .map(Some)
-        .map_err(ApiError::from)
+        .map_err(|error| ApiError::json_deserialize(provider, model, &payload, error))
 }

 #[cfg(test)]
@@ -276,4 +303,28 @@ mod tests {
            ))
        );
    }
+
+    #[test]
+    fn given_message_delta_frame_with_empty_usage_when_parsed_then_usage_defaults_to_zero() {
+        // given
+        let frame = concat!(
+            "event: message_delta\n",
+            "data: {\"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"end_turn\",\"stop_sequence\":null},\"usage\":{}}\n\n"
+        );
+
+        // when
+        let event = parse_frame(frame).expect("frame should parse");
+
+        // then
+        assert_eq!(
+            event,
+            Some(StreamEvent::MessageDelta(crate::types::MessageDeltaEvent {
+                delta: MessageDelta {
+                    stop_reason: Some("end_turn".to_string()),
+                    stop_sequence: None,
+                },
+                usage: Usage::default(),
+            }))
+        );
+    }
 }
@@ -1,7 +1,8 @@
+use runtime::{pricing_for_model, TokenUsage, UsageCostEstimate};
 use serde::{Deserialize, Serialize};
 use serde_json::Value;

-#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
+#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, Default)]
 pub struct MessageRequest {
    pub model: String,
    pub max_tokens: u32,
@@ -14,6 +15,22 @@ pub struct MessageRequest {
    pub tool_choice: Option<ToolChoice>,
    #[serde(default, skip_serializing_if = "std::ops::Not::not")]
    pub stream: bool,
+    /// OpenAI-compatible tuning parameters. Optional — omitted from payload when None.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub temperature: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub top_p: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub frequency_penalty: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub presence_penalty: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub stop: Option<Vec<String>>,
+    /// Reasoning effort level for OpenAI-compatible reasoning models (e.g. `o4-mini`).
+    /// Accepted values: `"low"`, `"medium"`, `"high"`. Omitted when `None`.
+    /// Silently ignored by backends that do not support it.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub reasoning_effort: Option<String>,
 }

 impl MessageRequest {
@@ -112,6 +129,7 @@ pub struct MessageResponse {
    pub stop_reason: Option<String>,
    #[serde(default)]
    pub stop_sequence: Option<String>,
+    #[serde(default)]
    pub usage: Usage,
    #[serde(default)]
    pub request_id: Option<String>,
@@ -146,20 +164,44 @@ pub enum OutputContentBlock {
    },
 }

-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Debug, Clone, Default, PartialEq, Eq, Serialize, Deserialize)]
 pub struct Usage {
+    #[serde(default)]
    pub input_tokens: u32,
    #[serde(default)]
    pub cache_creation_input_tokens: u32,
    #[serde(default)]
    pub cache_read_input_tokens: u32,
+    #[serde(default)]
    pub output_tokens: u32,
 }

 impl Usage {
    #[must_use]
    pub const fn total_tokens(&self) -> u32 {
-        self.input_tokens + self.output_tokens
+        self.input_tokens
+            + self.output_tokens
+            + self.cache_creation_input_tokens
+            + self.cache_read_input_tokens
+    }
+
+    #[must_use]
+    pub const fn token_usage(&self) -> TokenUsage {
+        TokenUsage {
+            input_tokens: self.input_tokens,
+            output_tokens: self.output_tokens,
+            cache_creation_input_tokens: self.cache_creation_input_tokens,
+            cache_read_input_tokens: self.cache_read_input_tokens,
+        }
+    }
+
+    #[must_use]
+    pub fn estimated_cost_usd(&self, model: &str) -> UsageCostEstimate {
+        let usage = self.token_usage();
+        pricing_for_model(model).map_or_else(
+            || usage.estimate_cost_usd(),
+            |pricing| usage.estimate_cost_usd_with_pricing(pricing),
+        )
    }
 }

@@ -171,6 +213,7 @@ pub struct MessageStartEvent {
 #[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
 pub struct MessageDeltaEvent {
    pub delta: MessageDelta,
+    #[serde(default)]
    pub usage: Usage,
 }

@@ -221,3 +264,47 @@ pub enum StreamEvent {
    ContentBlockStop(ContentBlockStopEvent),
    MessageStop(MessageStopEvent),
 }
+
+#[cfg(test)]
+mod tests {
+    use runtime::format_usd;
+
+    use super::{MessageResponse, Usage};
+
+    #[test]
+    fn usage_total_tokens_includes_cache_tokens() {
+        let usage = Usage {
+            input_tokens: 10,
+            cache_creation_input_tokens: 2,
+            cache_read_input_tokens: 3,
+            output_tokens: 4,
+        };
+
+        assert_eq!(usage.total_tokens(), 19);
+        assert_eq!(usage.token_usage().total_tokens(), 19);
+    }
+
+    #[test]
+    fn message_response_estimates_cost_from_model_usage() {
+        let response = MessageResponse {
+            id: "msg_cost".to_string(),
+            kind: "message".to_string(),
+            role: "assistant".to_string(),
+            content: Vec::new(),
+            model: "claude-sonnet-4-20250514".to_string(),
+            stop_reason: Some("end_turn".to_string()),
+            stop_sequence: None,
+            usage: Usage {
+                input_tokens: 1_000_000,
+                cache_creation_input_tokens: 100_000,
+                cache_read_input_tokens: 200_000,
+                output_tokens: 500_000,
+            },
+            request_id: None,
+        };
+
+        let cost = response.usage.estimated_cost_usd(&response.model);
+        assert_eq!(format_usd(cost.total_cost_usd()), "$54.6750");
+        assert_eq!(response.total_tokens(), 1_800_000);
+    }
+}
@@ -1,17 +1,27 @@
 use std::collections::HashMap;
 use std::sync::Arc;
+use std::sync::{Mutex as StdMutex, OnceLock};
 use std::time::Duration;

 use api::{
-    ApiClient, ApiError, AuthSource, ContentBlockDelta, ContentBlockDeltaEvent,
+    AnthropicClient, ApiClient, ApiError, AuthSource, ContentBlockDelta, ContentBlockDeltaEvent,
    ContentBlockStartEvent, InputContentBlock, InputMessage, MessageDeltaEvent, MessageRequest,
-    OutputContentBlock, ProviderClient, StreamEvent, ToolChoice, ToolDefinition,
+    OutputContentBlock, PromptCache, PromptCacheConfig, ProviderClient, StreamEvent, ToolChoice,
+    ToolDefinition,
 };
 use serde_json::json;
+use telemetry::{ClientIdentity, MemoryTelemetrySink, SessionTracer, TelemetryEvent};
 use tokio::io::{AsyncReadExt, AsyncWriteExt};
 use tokio::net::TcpListener;
 use tokio::sync::Mutex;

+fn env_lock() -> std::sync::MutexGuard<'static, ()> {
+    static LOCK: OnceLock<StdMutex<()>> = OnceLock::new();
+    LOCK.get_or_init(|| StdMutex::new(()))
+        .lock()
+        .unwrap_or_else(std::sync::PoisonError::into_inner)
+}
+
 #[tokio::test]
 async fn send_message_posts_json_and_parses_response() {
    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
@@ -20,8 +30,8 @@ async fn send_message_posts_json_and_parses_response() {
        "\"id\":\"msg_test\",",
        "\"type\":\"message\",",
        "\"role\":\"assistant\",",
-        "\"content\":[{\"type\":\"text\",\"text\":\"Hello from Claw\"}],",
-        "\"model\":\"claude-sonnet-4-6\",",
+        "\"content\":[{\"type\":\"text\",\"text\":\"Hello from Claude\"}],",
+        "\"model\":\"claude-3-7-sonnet-latest\",",
        "\"stop_reason\":\"end_turn\",",
        "\"stop_sequence\":null,",
        "\"usage\":{\"input_tokens\":12,\"output_tokens\":4},",
@@ -45,10 +55,12 @@ async fn send_message_posts_json_and_parses_response() {
    assert_eq!(response.id, "msg_test");
    assert_eq!(response.total_tokens(), 16);
    assert_eq!(response.request_id.as_deref(), Some("req_body_123"));
+    assert_eq!(response.usage.cache_creation_input_tokens, 0);
+    assert_eq!(response.usage.cache_read_input_tokens, 0);
    assert_eq!(
        response.content,
        vec![OutputContentBlock::Text {
-            text: "Hello from Claw".to_string(),
+            text: "Hello from Claude".to_string(),
        }]
    );

@@ -64,23 +76,258 @@ async fn send_message_posts_json_and_parses_response() {
        request.headers.get("authorization").map(String::as_str),
        Some("Bearer proxy-token")
    );
+    assert_eq!(
+        request.headers.get("anthropic-version").map(String::as_str),
+        Some("2023-06-01")
+    );
+    assert_eq!(
+        request.headers.get("user-agent").map(String::as_str),
+        Some("claude-code/0.1.0")
+    );
+    assert_eq!(
+        request.headers.get("anthropic-beta").map(String::as_str),
+        Some("claude-code-20250219,prompt-caching-scope-2026-01-05")
+    );
    let body: serde_json::Value =
        serde_json::from_str(&request.body).expect("request body should be json");
    assert_eq!(
        body.get("model").and_then(serde_json::Value::as_str),
-        Some("claude-sonnet-4-6")
+        Some("claude-3-7-sonnet-latest")
    );
    assert!(body.get("stream").is_none());
    assert_eq!(body["tools"][0]["name"], json!("get_weather"));
    assert_eq!(body["tool_choice"]["type"], json!("auto"));
+    assert!(
+        body.get("betas").is_none(),
+        "betas must travel via the anthropic-beta header, not the request body"
+    );
 }

 #[tokio::test]
+async fn send_message_blocks_oversized_requests_before_the_http_call() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let server = spawn_server(
+        state.clone(),
+        vec![http_response("200 OK", "application/json", "{}")],
+    )
+    .await;
+
+    let client = AnthropicClient::new("test-key").with_base_url(server.base_url());
+    let error = client
+        .send_message(&MessageRequest {
+            model: "claude-sonnet-4-6".to_string(),
+            max_tokens: 64_000,
+            messages: vec![InputMessage {
+                role: "user".to_string(),
+                content: vec![InputContentBlock::Text {
+                    text: "x".repeat(600_000),
+                }],
+            }],
+            system: Some("Keep the answer short.".to_string()),
+            tools: None,
+            tool_choice: None,
+            stream: false,
+            ..Default::default()
+        })
+        .await
+        .expect_err("oversized request should fail local context-window preflight");
+
+    assert!(matches!(error, ApiError::ContextWindowExceeded { .. }));
+    assert!(
+        state.lock().await.is_empty(),
+        "preflight failure should avoid any upstream HTTP request"
+    );
+}
+
+#[tokio::test]
+async fn send_message_applies_request_profile_and_records_telemetry() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let server = spawn_server(
+        state.clone(),
+        vec![http_response_with_headers(
+            "200 OK",
+            "application/json",
+            concat!(
+                "{",
+                "\"id\":\"msg_profile\",",
+                "\"type\":\"message\",",
+                "\"role\":\"assistant\",",
+                "\"content\":[{\"type\":\"text\",\"text\":\"ok\"}],",
+                "\"model\":\"claude-3-7-sonnet-latest\",",
+                "\"stop_reason\":\"end_turn\",",
+                "\"stop_sequence\":null,",
+                "\"usage\":{\"input_tokens\":1,\"cache_creation_input_tokens\":2,\"cache_read_input_tokens\":3,\"output_tokens\":1}",
+                "}"
+            ),
+            &[("request-id", "req_profile_123")],
+        )],
+    )
+    .await;
+    let sink = Arc::new(MemoryTelemetrySink::default());
+
+    let client = AnthropicClient::new("test-key")
+        .with_base_url(server.base_url())
+        .with_client_identity(ClientIdentity::new("claude-code", "9.9.9").with_runtime("rust-cli"))
+        .with_beta("tools-2026-04-01")
+        .with_extra_body_param("metadata", json!({"source": "clawd-code"}))
+        .with_session_tracer(SessionTracer::new("session-telemetry", sink.clone()));
+
+    let response = client
+        .send_message(&sample_request(false))
+        .await
+        .expect("request should succeed");
+
+    assert_eq!(response.request_id.as_deref(), Some("req_profile_123"));
+
+    let captured = state.lock().await;
+    let request = captured.first().expect("server should capture request");
+    assert_eq!(
+        request.headers.get("anthropic-beta").map(String::as_str),
+        Some("claude-code-20250219,prompt-caching-scope-2026-01-05,tools-2026-04-01")
+    );
+    assert_eq!(
+        request.headers.get("user-agent").map(String::as_str),
+        Some("claude-code/9.9.9")
+    );
+    let body: serde_json::Value =
+        serde_json::from_str(&request.body).expect("request body should be json");
+    assert_eq!(body["metadata"]["source"], json!("clawd-code"));
+    assert!(
+        body.get("betas").is_none(),
+        "betas must travel via the anthropic-beta header, not the request body"
+    );
+
+    let events = sink.events();
+    assert_eq!(events.len(), 6);
+    assert!(matches!(
+        &events[0],
+        TelemetryEvent::HttpRequestStarted {
+            session_id,
+            attempt: 1,
+            method,
+            path,
+            ..
+        } if session_id == "session-telemetry" && method == "POST" && path == "/v1/messages"
+    ));
+    assert!(matches!(
+        &events[1],
+        TelemetryEvent::SessionTrace(trace) if trace.name == "http_request_started"
+    ));
+    assert!(matches!(
+        &events[2],
+        TelemetryEvent::HttpRequestSucceeded {
+            request_id,
+            status: 200,
+            ..
+        } if request_id.as_deref() == Some("req_profile_123")
+    ));
+    assert!(matches!(
+        &events[3],
+        TelemetryEvent::SessionTrace(trace) if trace.name == "http_request_succeeded"
+    ));
+    assert!(matches!(
+        &events[4],
+        TelemetryEvent::Analytics(event)
+            if event.namespace == "api"
+                && event.action == "message_usage"
+                && event.properties.get("request_id") == Some(&json!("req_profile_123"))
+                && event.properties.get("total_tokens") == Some(&json!(7))
+                && event.properties.get("estimated_cost_usd") == Some(&json!("$0.0001"))
+    ));
+    assert!(matches!(
+        &events[5],
+        TelemetryEvent::SessionTrace(trace) if trace.name == "analytics"
+    ));
+}
+
+#[tokio::test]
+async fn send_message_parses_prompt_cache_token_usage_from_response() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let body = concat!(
+        "{",
+        "\"id\":\"msg_cache_tokens\",",
+        "\"type\":\"message\",",
+        "\"role\":\"assistant\",",
+        "\"content\":[{\"type\":\"text\",\"text\":\"Cache tokens\"}],",
+        "\"model\":\"claude-3-7-sonnet-latest\",",
+        "\"stop_reason\":\"end_turn\",",
+        "\"stop_sequence\":null,",
+        "\"usage\":{\"input_tokens\":12,\"cache_creation_input_tokens\":321,\"cache_read_input_tokens\":654,\"output_tokens\":4}",
+        "}"
+    );
+    let server = spawn_server(
+        state,
+        vec![http_response("200 OK", "application/json", body)],
+    )
+    .await;
+
+    let client = AnthropicClient::new("test-key").with_base_url(server.base_url());
+    let response = client
+        .send_message(&sample_request(false))
+        .await
+        .expect("request should succeed");
+
+    assert_eq!(response.usage.input_tokens, 12);
+    assert_eq!(response.usage.cache_creation_input_tokens, 321);
+    assert_eq!(response.usage.cache_read_input_tokens, 654);
+    assert_eq!(response.usage.output_tokens, 4);
+}
+
+#[tokio::test]
+async fn given_empty_usage_object_when_send_message_parses_response_then_usage_defaults_to_zero() {
+    // given
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let body = concat!(
+        "{",
+        "\"id\":\"msg_empty_usage\",",
+        "\"type\":\"message\",",
+        "\"role\":\"assistant\",",
+        "\"content\":[{\"type\":\"text\",\"text\":\"Hello from Claude\"}],",
+        "\"model\":\"claude-3-7-sonnet-latest\",",
+        "\"stop_reason\":\"end_turn\",",
+        "\"stop_sequence\":null,",
+        "\"usage\":{}",
+        "}"
+    );
+    let server = spawn_server(
+        state,
+        vec![http_response("200 OK", "application/json", body)],
+    )
+    .await;
+    let client = AnthropicClient::new("test-key").with_base_url(server.base_url());
+
+    // when
+    let response = client
+        .send_message(&sample_request(false))
+        .await
+        .expect("response with empty usage object should still parse");
+
+    // then
+    assert_eq!(response.id, "msg_empty_usage");
+    assert_eq!(response.total_tokens(), 0);
+    assert_eq!(response.usage.input_tokens, 0);
+    assert_eq!(response.usage.cache_creation_input_tokens, 0);
+    assert_eq!(response.usage.cache_read_input_tokens, 0);
+    assert_eq!(response.usage.output_tokens, 0);
+}
+
+#[tokio::test]
+#[allow(clippy::await_holding_lock)]
 async fn stream_message_parses_sse_events_with_tool_use() {
+    let _guard = env_lock();
+    let temp_root = std::env::temp_dir().join(format!(
+        "api-stream-cache-{}-{}",
+        std::process::id(),
+        std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .expect("time")
+            .as_nanos()
+    ));
+    std::env::set_var("CLAUDE_CONFIG_HOME", &temp_root);
    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
    let sse = concat!(
        "event: message_start\n",
-        "data: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_stream\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[],\"model\":\"claude-sonnet-4-6\",\"stop_reason\":null,\"stop_sequence\":null,\"usage\":{\"input_tokens\":8,\"output_tokens\":0}}}\n\n",
+        "data: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_stream\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[],\"model\":\"claude-3-7-sonnet-latest\",\"stop_reason\":null,\"stop_sequence\":null,\"usage\":{\"input_tokens\":8,\"cache_creation_input_tokens\":13,\"cache_read_input_tokens\":21,\"output_tokens\":0}}}\n\n",
        "event: content_block_start\n",
        "data: {\"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"tool_use\",\"id\":\"toolu_123\",\"name\":\"get_weather\",\"input\":{}}}\n\n",
        "event: content_block_delta\n",
@@ -88,7 +335,7 @@ async fn stream_message_parses_sse_events_with_tool_use() {
        "event: content_block_stop\n",
        "data: {\"type\":\"content_block_stop\",\"index\":0}\n\n",
        "event: message_delta\n",
-        "data: {\"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"tool_use\",\"stop_sequence\":null},\"usage\":{\"input_tokens\":8,\"output_tokens\":1}}\n\n",
+        "data: {\"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"tool_use\",\"stop_sequence\":null},\"usage\":{\"input_tokens\":8,\"cache_creation_input_tokens\":34,\"cache_read_input_tokens\":55,\"output_tokens\":1}}\n\n",
        "event: message_stop\n",
        "data: {\"type\":\"message_stop\"}\n\n",
        "data: [DONE]\n\n"
@@ -106,7 +353,8 @@ async fn stream_message_parses_sse_events_with_tool_use() {

    let client = ApiClient::new("test-key")
        .with_auth_token(Some("proxy-token".to_string()))
-        .with_base_url(server.base_url());
+        .with_base_url(server.base_url())
+        .with_prompt_cache(PromptCache::new("stream-session"));
    let mut stream = client
        .stream_message(&sample_request(false))
        .await
@@ -160,6 +408,20 @@ async fn stream_message_parses_sse_events_with_tool_use() {
    let captured = state.lock().await;
    let request = captured.first().expect("server should capture request");
    assert!(request.body.contains("\"stream\":true"));
+
+    let cache_stats = client
+        .prompt_cache_stats()
+        .expect("prompt cache stats should exist");
+    assert_eq!(cache_stats.tracked_requests, 1);
+    assert_eq!(cache_stats.last_cache_creation_input_tokens, Some(34));
+    assert_eq!(cache_stats.last_cache_read_input_tokens, Some(55));
+    assert_eq!(
+        cache_stats.last_cache_source.as_deref(),
+        Some("api-response")
+    );
+
+    std::fs::remove_dir_all(temp_root).expect("cleanup temp root");
+    std::env::remove_var("CLAUDE_CONFIG_HOME");
 }

 #[tokio::test]
@@ -176,7 +438,7 @@ async fn retries_retryable_failures_before_succeeding() {
            http_response(
                "200 OK",
                "application/json",
-                "{\"id\":\"msg_retry\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"Recovered\"}],\"model\":\"claude-sonnet-4-6\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"output_tokens\":2}}",
+                "{\"id\":\"msg_retry\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"Recovered\"}],\"model\":\"claude-3-7-sonnet-latest\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"output_tokens\":2}}",
            ),
        ],
    )
@@ -196,28 +458,28 @@ async fn retries_retryable_failures_before_succeeding() {
 }

 #[tokio::test]
-async fn provider_client_dispatches_api_requests() {
+async fn provider_client_dispatches_anthropic_requests() {
    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
    let server = spawn_server(
        state.clone(),
        vec![http_response(
            "200 OK",
            "application/json",
-            "{\"id\":\"msg_provider\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"Dispatched\"}],\"model\":\"claude-sonnet-4-6\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"output_tokens\":2}}",
+            "{\"id\":\"msg_provider\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"Dispatched\"}],\"model\":\"claude-3-7-sonnet-latest\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"output_tokens\":2}}",
        )],
    )
    .await;

-    let client = ProviderClient::from_model_with_default_auth(
+    let client = ProviderClient::from_model_with_anthropic_auth(
        "claude-sonnet-4-6",
        Some(AuthSource::ApiKey("test-key".to_string())),
    )
-    .expect("api provider client should be constructed");
+    .expect("anthropic provider client should be constructed");
    let client = match client {
-        ProviderClient::ClawApi(client) => {
-            ProviderClient::ClawApi(client.with_base_url(server.base_url()))
+        ProviderClient::Anthropic(client) => {
+            ProviderClient::Anthropic(client.with_base_url(server.base_url()))
        }
-        other => panic!("expected default provider, got {other:?}"),
+        other => panic!("expected anthropic provider, got {other:?}"),
    };

    let response = client
@@ -284,13 +546,194 @@ async fn surfaces_retry_exhaustion_for_persistent_retryable_errors() {
    }
 }

+#[tokio::test]
+async fn retries_multiple_retryable_failures_with_exponential_backoff_and_jitter() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let server = spawn_server(
+        state.clone(),
+        vec![
+            http_response(
+                "429 Too Many Requests",
+                "application/json",
+                "{\"type\":\"error\",\"error\":{\"type\":\"rate_limit_error\",\"message\":\"slow down\"}}",
+            ),
+            http_response(
+                "500 Internal Server Error",
+                "application/json",
+                "{\"type\":\"error\",\"error\":{\"type\":\"api_error\",\"message\":\"boom\"}}",
+            ),
+            http_response(
+                "503 Service Unavailable",
+                "application/json",
+                "{\"type\":\"error\",\"error\":{\"type\":\"overloaded_error\",\"message\":\"busy\"}}",
+            ),
+            http_response(
+                "429 Too Many Requests",
+                "application/json",
+                "{\"type\":\"error\",\"error\":{\"type\":\"rate_limit_error\",\"message\":\"slow down again\"}}",
+            ),
+            http_response(
+                "503 Service Unavailable",
+                "application/json",
+                "{\"type\":\"error\",\"error\":{\"type\":\"overloaded_error\",\"message\":\"still busy\"}}",
+            ),
+            http_response(
+                "200 OK",
+                "application/json",
+                "{\"id\":\"msg_exp_retry\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"Recovered after 5\"}],\"model\":\"claude-3-7-sonnet-latest\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"output_tokens\":2}}",
+            ),
+        ],
+    )
+    .await;
+
+    let client = ApiClient::new("test-key")
+        .with_base_url(server.base_url())
+        .with_retry_policy(8, Duration::from_millis(1), Duration::from_millis(4));
+    let started_at = std::time::Instant::now();
+
+    let response = client
+        .send_message(&sample_request(false))
+        .await
+        .expect("8-retry policy should absorb 5 retryable failures");
+
+    let elapsed = started_at.elapsed();
+    assert_eq!(response.total_tokens(), 5);
+    assert_eq!(
+        state.lock().await.len(),
+        6,
+        "client should issue 1 original + 5 retry requests before the 200"
+    );
+    // Jittered sleeps are bounded by 2 * max_backoff per retry (base + jitter),
+    // so 5 sleeps fit comfortably below this upper bound with generous slack.
+    assert!(
+        elapsed < Duration::from_secs(5),
+        "retries should complete promptly, took {elapsed:?}"
+    );
+}
+
+#[tokio::test]
+#[allow(clippy::await_holding_lock)]
+async fn send_message_reuses_recent_completion_cache_entries() {
+    let _guard = env_lock();
+    let temp_root = std::env::temp_dir().join(format!(
+        "api-prompt-cache-{}-{}",
+        std::process::id(),
+        std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .expect("time")
+            .as_nanos()
+    ));
+    std::env::set_var("CLAUDE_CONFIG_HOME", &temp_root);
+
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let server = spawn_server(
+        state.clone(),
+        vec![http_response(
+            "200 OK",
+            "application/json",
+            "{\"id\":\"msg_cached\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"Cached once\"}],\"model\":\"claude-3-7-sonnet-latest\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"cache_creation_input_tokens\":5,\"cache_read_input_tokens\":4000,\"output_tokens\":2}}",
+        )],
+    )
+    .await;
+
+    let client = AnthropicClient::new("test-key")
+        .with_base_url(server.base_url())
+        .with_prompt_cache(PromptCache::new("integration-session"));
+
+    let first = client
+        .send_message(&sample_request(false))
+        .await
+        .expect("first request should succeed");
+    let second = client
+        .send_message(&sample_request(false))
+        .await
+        .expect("second request should reuse cache");
+
+    assert_eq!(first.content, second.content);
+    assert_eq!(state.lock().await.len(), 1);
+
+    let cache_stats = client
+        .prompt_cache_stats()
+        .expect("prompt cache stats should exist");
+    assert_eq!(cache_stats.completion_cache_hits, 1);
+    assert_eq!(cache_stats.completion_cache_misses, 1);
+    assert_eq!(cache_stats.completion_cache_writes, 1);
+
+    std::fs::remove_dir_all(temp_root).expect("cleanup temp root");
+    std::env::remove_var("CLAUDE_CONFIG_HOME");
+}
+
+#[tokio::test]
+#[allow(clippy::await_holding_lock)]
+async fn send_message_tracks_unexpected_prompt_cache_breaks() {
+    let _guard = env_lock();
+    let temp_root = std::env::temp_dir().join(format!(
+        "api-prompt-break-{}-{}",
+        std::process::id(),
+        std::time::SystemTime::now()
+            .duration_since(std::time::UNIX_EPOCH)
+            .expect("time")
+            .as_nanos()
+    ));
+    std::env::set_var("CLAUDE_CONFIG_HOME", &temp_root);
+
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let server = spawn_server(
+        state,
+        vec![
+            http_response(
+                "200 OK",
+                "application/json",
+                "{\"id\":\"msg_one\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"One\"}],\"model\":\"claude-3-7-sonnet-latest\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"cache_creation_input_tokens\":5,\"cache_read_input_tokens\":6000,\"output_tokens\":2}}",
+            ),
+            http_response(
+                "200 OK",
+                "application/json",
+                "{\"id\":\"msg_two\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"text\",\"text\":\"Two\"}],\"model\":\"claude-3-7-sonnet-latest\",\"stop_reason\":\"end_turn\",\"stop_sequence\":null,\"usage\":{\"input_tokens\":3,\"cache_creation_input_tokens\":0,\"cache_read_input_tokens\":1000,\"output_tokens\":2}}",
+            ),
+        ],
+    )
+    .await;
+
+    let request = sample_request(false);
+    let client = AnthropicClient::new("test-key")
+        .with_base_url(server.base_url())
+        .with_prompt_cache(PromptCache::with_config(PromptCacheConfig {
+            session_id: "break-session".to_string(),
+            completion_ttl: Duration::from_secs(0),
+            ..PromptCacheConfig::default()
+        }));
+
+    client
+        .send_message(&request)
+        .await
+        .expect("first response should succeed");
+    client
+        .send_message(&request)
+        .await
+        .expect("second response should succeed");
+
+    let cache_stats = client
+        .prompt_cache_stats()
+        .expect("prompt cache stats should exist");
+    assert_eq!(cache_stats.unexpected_cache_breaks, 1);
+    assert_eq!(
+        cache_stats.last_break_reason.as_deref(),
+        Some("cache read tokens dropped while prompt fingerprint remained stable")
+    );
+
+    std::fs::remove_dir_all(temp_root).expect("cleanup temp root");
+    std::env::remove_var("CLAUDE_CONFIG_HOME");
+}
+
 #[tokio::test]
 #[ignore = "requires ANTHROPIC_API_KEY and network access"]
 async fn live_stream_smoke_test() {
    let client = ApiClient::from_env().expect("ANTHROPIC_API_KEY must be set");
    let mut stream = client
        .stream_message(&MessageRequest {
-            model: std::env::var("CLAW_MODEL").unwrap_or_else(|_| "claude-sonnet-4-6".to_string()),
+            model: std::env::var("ANTHROPIC_MODEL")
+                .unwrap_or_else(|_| "claude-3-7-sonnet-latest".to_string()),
            max_tokens: 32,
            messages: vec![InputMessage::user_text(
                "Reply with exactly: hello from rust",
@@ -299,6 +742,7 @@ async fn live_stream_smoke_test() {
            tools: None,
            tool_choice: None,
            stream: false,
+            ..Default::default()
        })
        .await
        .expect("live stream should start");
@@ -450,7 +894,7 @@ fn http_response_with_headers(

 fn sample_request(stream: bool) -> MessageRequest {
    MessageRequest {
-        model: "claude-sonnet-4-6".to_string(),
+        model: "claude-3-7-sonnet-latest".to_string(),
        max_tokens: 64,
        messages: vec![InputMessage {
            role: "user".to_string(),
@@ -479,5 +923,6 @@ fn sample_request(stream: bool) -> MessageRequest {
        }]),
        tool_choice: Some(ToolChoice::Auto),
        stream,
+        ..Default::default()
    }
 }
@@ -4,9 +4,10 @@ use std::sync::Arc;
 use std::sync::{Mutex as StdMutex, OnceLock};

 use api::{
-    ContentBlockDelta, ContentBlockDeltaEvent, ContentBlockStartEvent, ContentBlockStopEvent,
-    InputContentBlock, InputMessage, MessageRequest, OpenAiCompatClient, OpenAiCompatConfig,
-    OutputContentBlock, ProviderClient, StreamEvent, ToolChoice, ToolDefinition,
+    ApiError, ContentBlockDelta, ContentBlockDeltaEvent, ContentBlockStartEvent,
+    ContentBlockStopEvent, InputContentBlock, InputMessage, MessageDeltaEvent, MessageRequest,
+    OpenAiCompatClient, OpenAiCompatConfig, OutputContentBlock, ProviderClient, StreamEvent,
+    ToolChoice, ToolDefinition,
 };
 use serde_json::json;
 use tokio::io::{AsyncReadExt, AsyncWriteExt};
@@ -62,6 +63,43 @@ async fn send_message_uses_openai_compatible_endpoint_and_auth() {
    assert_eq!(body["tools"][0]["type"], json!("function"));
 }

+#[tokio::test]
+async fn send_message_blocks_oversized_xai_requests_before_the_http_call() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let server = spawn_server(
+        state.clone(),
+        vec![http_response("200 OK", "application/json", "{}")],
+    )
+    .await;
+
+    let client = OpenAiCompatClient::new("xai-test-key", OpenAiCompatConfig::xai())
+        .with_base_url(server.base_url());
+    let error = client
+        .send_message(&MessageRequest {
+            model: "grok-3".to_string(),
+            max_tokens: 64_000,
+            messages: vec![InputMessage {
+                role: "user".to_string(),
+                content: vec![InputContentBlock::Text {
+                    text: "x".repeat(300_000),
+                }],
+            }],
+            system: Some("Keep the answer short.".to_string()),
+            tools: None,
+            tool_choice: None,
+            stream: false,
+            ..Default::default()
+        })
+        .await
+        .expect_err("oversized request should fail local context-window preflight");
+
+    assert!(matches!(error, ApiError::ContextWindowExceeded { .. }));
+    assert!(
+        state.lock().await.is_empty(),
+        "preflight failure should avoid any upstream HTTP request"
+    );
+}
+
 #[tokio::test]
 async fn send_message_accepts_full_chat_completions_endpoint_override() {
    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
@@ -195,6 +233,83 @@ async fn stream_message_normalizes_text_and_multiple_tool_calls() {
    assert!(request.body.contains("\"stream\":true"));
 }

+#[allow(clippy::await_holding_lock)]
+#[tokio::test]
+async fn openai_streaming_requests_opt_into_usage_chunks() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let sse = concat!(
+        "data: {\"id\":\"chatcmpl_openai_stream\",\"model\":\"gpt-5\",\"choices\":[{\"delta\":{\"content\":\"Hi\"}}]}\n\n",
+        "data: {\"id\":\"chatcmpl_openai_stream\",\"choices\":[{\"delta\":{},\"finish_reason\":\"stop\"}]}\n\n",
+        "data: {\"id\":\"chatcmpl_openai_stream\",\"choices\":[],\"usage\":{\"prompt_tokens\":9,\"completion_tokens\":4}}\n\n",
+        "data: [DONE]\n\n"
+    );
+    let server = spawn_server(
+        state.clone(),
+        vec![http_response_with_headers(
+            "200 OK",
+            "text/event-stream",
+            sse,
+            &[("x-request-id", "req_openai_stream")],
+        )],
+    )
+    .await;
+
+    let client = OpenAiCompatClient::new("openai-test-key", OpenAiCompatConfig::openai())
+        .with_base_url(server.base_url());
+    let mut stream = client
+        .stream_message(&sample_request(false))
+        .await
+        .expect("stream should start");
+
+    assert_eq!(stream.request_id(), Some("req_openai_stream"));
+
+    let mut events = Vec::new();
+    while let Some(event) = stream.next_event().await.expect("event should parse") {
+        events.push(event);
+    }
+
+    assert!(matches!(events[0], StreamEvent::MessageStart(_)));
+    assert!(matches!(
+        events[1],
+        StreamEvent::ContentBlockStart(ContentBlockStartEvent {
+            content_block: OutputContentBlock::Text { .. },
+            ..
+        })
+    ));
+    assert!(matches!(
+        events[2],
+        StreamEvent::ContentBlockDelta(ContentBlockDeltaEvent {
+            delta: ContentBlockDelta::TextDelta { .. },
+            ..
+        })
+    ));
+    assert!(matches!(
+        events[3],
+        StreamEvent::ContentBlockStop(ContentBlockStopEvent { index: 0 })
+    ));
+    assert!(matches!(
+        events[4],
+        StreamEvent::MessageDelta(MessageDeltaEvent { .. })
+    ));
+    assert!(matches!(events[5], StreamEvent::MessageStop(_)));
+
+    match &events[4] {
+        StreamEvent::MessageDelta(MessageDeltaEvent { usage, .. }) => {
+            assert_eq!(usage.input_tokens, 9);
+            assert_eq!(usage.output_tokens, 4);
+        }
+        other => panic!("expected message delta, got {other:?}"),
+    }
+
+    let captured = state.lock().await;
+    let request = captured.first().expect("captured request");
+    assert_eq!(request.path, "/chat/completions");
+    let body: serde_json::Value = serde_json::from_str(&request.body).expect("json body");
+    assert_eq!(body["stream"], json!(true));
+    assert_eq!(body["stream_options"], json!({"include_usage": true}));
+}
+
+#[allow(clippy::await_holding_lock)]
 #[tokio::test]
 async fn provider_client_dispatches_xai_requests_from_env() {
    let _lock = env_lock();
@@ -382,6 +497,7 @@ fn sample_request(stream: bool) -> MessageRequest {
        }]),
        tool_choice: Some(ToolChoice::Auto),
        stream,
+        ..Default::default()
    }
 }

@@ -389,7 +505,7 @@ fn env_lock() -> std::sync::MutexGuard<'static, ()> {
    static LOCK: OnceLock<StdMutex<()>> = OnceLock::new();
    LOCK.get_or_init(|| StdMutex::new(()))
        .lock()
-        .unwrap_or_else(|poisoned| poisoned.into_inner())
+        .unwrap_or_else(std::sync::PoisonError::into_inner)
 }

 struct ScopedEnvVar {
@@ -22,7 +22,9 @@ fn provider_client_reports_missing_xai_credentials_for_grok_models() {
        .expect_err("grok requests without XAI_API_KEY should fail fast");

    match error {
-        ApiError::MissingCredentials { provider, env_vars } => {
+        ApiError::MissingCredentials {
+            provider, env_vars, ..
+        } => {
            assert_eq!(provider, "xAI");
            assert_eq!(env_vars, &["XAI_API_KEY"]);
        }
@@ -31,18 +33,18 @@ fn provider_client_reports_missing_xai_credentials_for_grok_models() {
 }

 #[test]
-fn provider_client_uses_explicit_auth_without_env_lookup() {
+fn provider_client_uses_explicit_anthropic_auth_without_env_lookup() {
    let _lock = env_lock();
-    let _api_key = EnvVarGuard::set("ANTHROPIC_API_KEY", None);
-    let _auth_token = EnvVarGuard::set("ANTHROPIC_AUTH_TOKEN", None);
+    let _anthropic_api_key = EnvVarGuard::set("ANTHROPIC_API_KEY", None);
+    let _anthropic_auth_token = EnvVarGuard::set("ANTHROPIC_AUTH_TOKEN", None);

-    let client = ProviderClient::from_model_with_default_auth(
+    let client = ProviderClient::from_model_with_anthropic_auth(
        "claude-sonnet-4-6",
-        Some(AuthSource::ApiKey("claw-test-key".to_string())),
+        Some(AuthSource::ApiKey("anthropic-test-key".to_string())),
    )
-    .expect("explicit auth should avoid env lookup");
+    .expect("explicit anthropic auth should avoid env lookup");

-    assert_eq!(client.provider_kind(), ProviderKind::ClawApi);
+    assert_eq!(client.provider_kind(), ProviderKind::Anthropic);
 }

 #[test]
@@ -57,7 +59,7 @@ fn env_lock() -> std::sync::MutexGuard<'static, ()> {
    static LOCK: OnceLock<Mutex<()>> = OnceLock::new();
    LOCK.get_or_init(|| Mutex::new(()))
        .lock()
-        .unwrap_or_else(|poisoned| poisoned.into_inner())
+        .unwrap_or_else(std::sync::PoisonError::into_inner)
 }

 struct EnvVarGuard {
@@ -0,0 +1,173 @@
+use std::ffi::OsString;
+use std::sync::{Mutex, OnceLock};
+
+use api::{build_http_client_with, ProxyConfig};
+
+fn env_lock() -> std::sync::MutexGuard<'static, ()> {
+    static LOCK: OnceLock<Mutex<()>> = OnceLock::new();
+    LOCK.get_or_init(|| Mutex::new(()))
+        .lock()
+        .unwrap_or_else(std::sync::PoisonError::into_inner)
+}
+
+struct EnvVarGuard {
+    key: &'static str,
+    original: Option<OsString>,
+}
+
+impl EnvVarGuard {
+    fn set(key: &'static str, value: Option<&str>) -> Self {
+        let original = std::env::var_os(key);
+        match value {
+            Some(value) => std::env::set_var(key, value),
+            None => std::env::remove_var(key),
+        }
+        Self { key, original }
+    }
+}
+
+impl Drop for EnvVarGuard {
+    fn drop(&mut self) {
+        match &self.original {
+            Some(value) => std::env::set_var(self.key, value),
+            None => std::env::remove_var(self.key),
+        }
+    }
+}
+
+#[test]
+fn proxy_config_from_env_reads_uppercase_proxy_vars() {
+    // given
+    let _lock = env_lock();
+    let _http = EnvVarGuard::set("HTTP_PROXY", Some("http://proxy.corp:3128"));
+    let _https = EnvVarGuard::set("HTTPS_PROXY", Some("http://secure.corp:3129"));
+    let _no = EnvVarGuard::set("NO_PROXY", Some("localhost,127.0.0.1"));
+    let _http_lower = EnvVarGuard::set("http_proxy", None);
+    let _https_lower = EnvVarGuard::set("https_proxy", None);
+    let _no_lower = EnvVarGuard::set("no_proxy", None);
+
+    // when
+    let config = ProxyConfig::from_env();
+
+    // then
+    assert_eq!(config.http_proxy.as_deref(), Some("http://proxy.corp:3128"));
+    assert_eq!(
+        config.https_proxy.as_deref(),
+        Some("http://secure.corp:3129")
+    );
+    assert_eq!(config.no_proxy.as_deref(), Some("localhost,127.0.0.1"));
+    assert!(config.proxy_url.is_none());
+    assert!(!config.is_empty());
+}
+
+#[test]
+fn proxy_config_from_env_reads_lowercase_proxy_vars() {
+    // given
+    let _lock = env_lock();
+    let _http = EnvVarGuard::set("HTTP_PROXY", None);
+    let _https = EnvVarGuard::set("HTTPS_PROXY", None);
+    let _no = EnvVarGuard::set("NO_PROXY", None);
+    let _http_lower = EnvVarGuard::set("http_proxy", Some("http://lower.corp:3128"));
+    let _https_lower = EnvVarGuard::set("https_proxy", Some("http://lower-secure.corp:3129"));
+    let _no_lower = EnvVarGuard::set("no_proxy", Some(".internal"));
+
+    // when
+    let config = ProxyConfig::from_env();
+
+    // then
+    assert_eq!(config.http_proxy.as_deref(), Some("http://lower.corp:3128"));
+    assert_eq!(
+        config.https_proxy.as_deref(),
+        Some("http://lower-secure.corp:3129")
+    );
+    assert_eq!(config.no_proxy.as_deref(), Some(".internal"));
+    assert!(!config.is_empty());
+}
+
+#[test]
+fn proxy_config_from_env_is_empty_when_no_vars_set() {
+    // given
+    let _lock = env_lock();
+    let _http = EnvVarGuard::set("HTTP_PROXY", None);
+    let _https = EnvVarGuard::set("HTTPS_PROXY", None);
+    let _no = EnvVarGuard::set("NO_PROXY", None);
+    let _http_lower = EnvVarGuard::set("http_proxy", None);
+    let _https_lower = EnvVarGuard::set("https_proxy", None);
+    let _no_lower = EnvVarGuard::set("no_proxy", None);
+
+    // when
+    let config = ProxyConfig::from_env();
+
+    // then
+    assert!(config.is_empty());
+    assert!(config.http_proxy.is_none());
+    assert!(config.https_proxy.is_none());
+    assert!(config.no_proxy.is_none());
+}
+
+#[test]
+fn proxy_config_from_env_treats_empty_values_as_unset() {
+    // given
+    let _lock = env_lock();
+    let _http = EnvVarGuard::set("HTTP_PROXY", Some(""));
+    let _https = EnvVarGuard::set("HTTPS_PROXY", Some(""));
+    let _http_lower = EnvVarGuard::set("http_proxy", Some(""));
+    let _https_lower = EnvVarGuard::set("https_proxy", Some(""));
+    let _no = EnvVarGuard::set("NO_PROXY", Some(""));
+    let _no_lower = EnvVarGuard::set("no_proxy", Some(""));
+
+    // when
+    let config = ProxyConfig::from_env();
+
+    // then
+    assert!(config.is_empty());
+}
+
+#[test]
+fn build_client_with_env_proxy_config_succeeds() {
+    // given
+    let _lock = env_lock();
+    let _http = EnvVarGuard::set("HTTP_PROXY", Some("http://proxy.corp:3128"));
+    let _https = EnvVarGuard::set("HTTPS_PROXY", Some("http://secure.corp:3129"));
+    let _no = EnvVarGuard::set("NO_PROXY", Some("localhost"));
+    let _http_lower = EnvVarGuard::set("http_proxy", None);
+    let _https_lower = EnvVarGuard::set("https_proxy", None);
+    let _no_lower = EnvVarGuard::set("no_proxy", None);
+    let config = ProxyConfig::from_env();
+
+    // when
+    let result = build_http_client_with(&config);
+
+    // then
+    assert!(result.is_ok());
+}
+
+#[test]
+fn build_client_with_proxy_url_config_succeeds() {
+    // given
+    let config = ProxyConfig::from_proxy_url("http://unified.corp:3128");
+
+    // when
+    let result = build_http_client_with(&config);
+
+    // then
+    assert!(result.is_ok());
+}
+
+#[test]
+fn proxy_config_from_env_prefers_uppercase_over_lowercase() {
+    // given
+    let _lock = env_lock();
+    let _http_upper = EnvVarGuard::set("HTTP_PROXY", Some("http://upper.corp:3128"));
+    let _http_lower = EnvVarGuard::set("http_proxy", Some("http://lower.corp:3128"));
+    let _https = EnvVarGuard::set("HTTPS_PROXY", None);
+    let _https_lower = EnvVarGuard::set("https_proxy", None);
+    let _no = EnvVarGuard::set("NO_PROXY", None);
+    let _no_lower = EnvVarGuard::set("no_proxy", None);
+
+    // when
+    let config = ProxyConfig::from_env();
+
+    // then
+    assert_eq!(config.http_proxy.as_deref(), Some("http://upper.corp:3128"));
+}
@@ -22,6 +22,7 @@ serde_json.workspace = true
 syntect = "5"
 tokio = { version = "1", features = ["rt-multi-thread", "time"] }
 tools = { path = "../tools" }
+dotenvy = "0.15"

 [lints]
 workspace = true
@@ -0,0 +1,58 @@
+# Claw CLI 模块 (claw-cli)
+
+本模块实现了 Claw 应用程序的主要命令行界面 (CLI)。它提供了交互式的 REPL 环境和非交互式的命令执行功能。
+
+## 概览
+
+`claw-cli` 模块是整个项目的“胶水”，负责编排 `runtime`、`api`、`tools` 和 `plugins` 模块之间的交互。它捕获用户输入，管理应用程序状态，并以用户友好的格式渲染 AI 响应。
+
+## 关键特性
+
+- **交互式 REPL**：一个功能齐全的 Read-Eval-Print Loop，用于与 AI 对话，支持：
+  - 通过 `rustyline` 实现多行输入和命令历史。
+  - 实时的 Markdown 流式显示和代码语法高亮。
+  - 为耗时的工具调用显示动态的加载动画 (Spinner) 和进度指示器。
+- **子命令**：
+  - `prompt`：运行单次 Prompt 并退出（单次模式）。
+  - `login`/`logout`：处理与 Claw 平台的 OAuth 身份验证。
+  - `init`：初始化新项目/仓库以配合 Claw 使用。
+  - `resume`：恢复并继续之前的对话会话。
+  - `agents`/`skills`：管理并发现可用的智能体和技能。
+- **权限管理**：对工具执行权限的细粒度控制：
+  - `read-only`：安全的分析模式。
+  - `workspace-write`：允许在当前工作区内进行修改。
+  - `danger-full-access`：高级任务的无限制访问。
+- **OAuth 流程**：集成了本地 HTTP 服务器，无缝处理用户机器上的 OAuth 回调重定向。
+
+## 实现逻辑
+
+### 核心模块
+
+- **`main.rs`**: 主要入口点。负责解析命令行参数、初始化环境，并调度到适当的操作（REPL 或子命令）。
+- **`render/`**: 包含 `TerminalRenderer` 和 Markdown 流渲染逻辑。使用 `syntect` 进行语法高亮，并使用 `crossterm` 进行终端操作。
+- **`input/`**: 处理用户输入捕获，包括对多行 Prompt 和斜杠命令 (Slash Command) 的特殊处理。
+- **`init.rs`**: 处理项目级初始化和仓库设置。
+
+### 交互循环流程
+
+1. CLI 初始化 `ConversationRuntime` 并加载项目上下文。
+2. 使用 `rustyline` 进入捕获用户输入的循环。
+3. 检查用户输入是否为“斜杠命令”（如 `/compact`、`/model`）。
+4. 普通 Prompt 通过 `runtime` 发送给 AI。
+5. AI 事件（文本增量、工具调用）逐步渲染到终端。
+6. 会话定期保存，以便将来可以恢复。
+
+## 使用方法
+
+主二进制程序名为 `claw`。
+
+```bash
+# 启动交互式 REPL
+claw
+
+# 运行单次 Prompt
+claw prompt "解释一下这个项目的架构"
+
+# 登录服务
+claw login
+```
@@ -16,7 +16,7 @@ use std::thread;
 use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};

 use api::{
-    resolve_startup_auth_source, AuthSource, ClawApiClient, ContentBlockDelta, InputContentBlock,
+    resolve_startup_auth_source, AnthropicClient, AuthSource, ContentBlockDelta, InputContentBlock,
    InputMessage, MessageRequest, MessageResponse, OutputContentBlock,
    StreamEvent as ApiStreamEvent, ToolChoice, ToolDefinition, ToolResultContentBlock,
 };
@@ -59,6 +59,7 @@ const INTERNAL_PROGRESS_HEARTBEAT_INTERVAL: Duration = Duration::from_secs(3);
 type AllowedToolSet = BTreeSet<String>;

 fn main() {
+    dotenvy::dotenv().ok();
    if let Err(error) = run() {
        eprintln!("{}", render_cli_error(&error.to_string()));
        std::process::exit(1);
@@ -171,7 +172,7 @@ impl CliOutputFormat {

 #[allow(clippy::too_many_lines)]
 fn parse_args(args: &[String]) -> Result<CliAction, String> {
-    let mut model = DEFAULT_MODEL.to_string();
+    let mut model = env::var("CLAW_MODEL").unwrap_or_else(|_| DEFAULT_MODEL.to_string());
    let mut output_format = CliOutputFormat::Text;
    let mut permission_mode = default_permission_mode();
    let mut wants_version = false;
@@ -328,7 +329,7 @@ fn join_optional_args(args: &[String]) -> Option<String> {

 fn parse_direct_slash_cli_action(rest: &[String]) -> Result<CliAction, String> {
    let raw = rest.join(" ");
-    match SlashCommand::parse(&raw) {
+    match SlashCommand::parse(&raw).map_err(|e| e.to_string())? {
        Some(SlashCommand::Help) => Ok(CliAction::Help),
        Some(SlashCommand::Agents { args }) => Ok(CliAction::Agents { args }),
        Some(SlashCommand::Skills { args }) => Ok(CliAction::Skills { args }),
@@ -483,7 +484,7 @@ fn dump_manifests() {
 }

 fn print_bootstrap_plan() {
-    for phase in runtime::BootstrapPlan::claw_default().phases() {
+    for phase in runtime::BootstrapPlan::claude_code_default().phases() {
        println!("- {phase:?}");
    }
 }
@@ -540,7 +541,7 @@ fn run_login() -> Result<(), Box<dyn std::error::Error>> {
        return Err(io::Error::new(io::ErrorKind::InvalidData, "oauth state mismatch").into());
    }

-    let client = ClawApiClient::from_auth(AuthSource::None).with_base_url(api::read_base_url());
+    let client = AnthropicClient::from_auth(AuthSource::None).with_base_url(api::read_base_url());
    let exchange_request =
        OAuthTokenExchangeRequest::from_config(oauth, code, state, pkce.verifier, redirect_uri);
    let runtime = tokio::runtime::Runtime::new()?;
@@ -649,7 +650,7 @@ fn resume_session(session_path: &Path, commands: &[String]) {

    let mut session = session;
    for raw_command in commands {
-        let Some(command) = SlashCommand::parse(raw_command) else {
+        let Ok(Some(command)) = SlashCommand::parse(raw_command) else {
            eprintln!("unsupported resumed command: {raw_command}");
            std::process::exit(2);
        };
@@ -986,8 +987,6 @@ fn run_resume_command(
        }
        SlashCommand::Bughunter { .. }
        | SlashCommand::Branch { .. }
-        | SlashCommand::Worktree { .. }
-        | SlashCommand::CommitPushPr { .. }
        | SlashCommand::Commit
        | SlashCommand::Pr { .. }
        | SlashCommand::Issue { .. }
@@ -999,7 +998,7 @@ fn run_resume_command(
        | SlashCommand::Permissions { .. }
        | SlashCommand::Session { .. }
        | SlashCommand::Plugins { .. }
-        | SlashCommand::Unknown(_) => Err("unsupported resumed slash command".into()),
+        | _ => Err("unsupported resumed slash command".into()),
    }
 }

@@ -1023,7 +1022,7 @@ fn run_repl(
                    cli.persist_session()?;
                    break;
                }
-                if let Some(command) = SlashCommand::parse(trimmed) {
+                if let Ok(Some(command)) = SlashCommand::parse(trimmed) {
                    if cli.handle_repl_command(command)? {
                        cli.persist_session()?;
                    }
@@ -1335,24 +1334,14 @@ impl LiveCli {
                );
                false
            }
-            SlashCommand::Worktree { .. } => {
-                eprintln!(
-                    "{}",
-                    render_mode_unavailable("worktree", "git worktree commands")
-                );
-                false
-            }
-            SlashCommand::CommitPushPr { .. } => {
-                eprintln!(
-                    "{}",
-                    render_mode_unavailable("commit-push-pr", "commit + push + PR automation")
-                );
-                false
-            }
            SlashCommand::Unknown(name) => {
                eprintln!("{}", render_unknown_repl_command(&name));
                false
            }
+            _ => {
+                eprintln!("command not available in this mode");
+                false
+            }
        })
    }

@@ -2988,7 +2977,7 @@ fn build_runtime(
        CliToolExecutor::new(allowed_tools.clone(), emit_output, tool_registry.clone()),
        permission_policy(permission_mode, &tool_registry),
        system_prompt,
-        feature_config,
+        &feature_config,
    ))
 }

@@ -3040,7 +3029,7 @@ impl runtime::PermissionPrompter for CliPermissionPrompter {

 struct DefaultRuntimeClient {
    runtime: tokio::runtime::Runtime,
-    client: ClawApiClient,
+    client: AnthropicClient,
    model: String,
    enable_tools: bool,
    emit_output: bool,
@@ -3060,7 +3049,7 @@ impl DefaultRuntimeClient {
    ) -> Result<Self, Box<dyn std::error::Error>> {
        Ok(Self {
            runtime: tokio::runtime::Runtime::new()?,
-            client: ClawApiClient::from_auth(resolve_cli_auth_source()?)
+            client: AnthropicClient::from_auth(resolve_cli_auth_source()?)
                .with_base_url(api::read_base_url()),
            model,
            enable_tools,
@@ -3098,6 +3087,12 @@ impl ApiClient for DefaultRuntimeClient {
                .then(|| filter_tool_specs(&self.tool_registry, self.allowed_tools.as_ref())),
            tool_choice: self.enable_tools.then_some(ToolChoice::Auto),
            stream: true,
+            temperature: None,
+            top_p: None,
+            frequency_penalty: None,
+            presence_penalty: None,
+            stop: None,
+            reasoning_effort: None,
        };

        self.runtime.block_on(async {
@@ -3158,8 +3153,17 @@ impl ApiClient for DefaultRuntimeClient {
                                input.push_str(&partial_json);
                            }
                        }
-                        ContentBlockDelta::ThinkingDelta { .. }
-                        | ContentBlockDelta::SignatureDelta { .. } => {}
+                        ContentBlockDelta::ThinkingDelta { thinking } => {
+                            if !thinking.is_empty() {
+                                if self.emit_output {
+                                    // Use a dimmed style for thinking blocks
+                                    write!(out, "\x1b[2m{thinking}\x1b[0m")
+                                        .and_then(|()| out.flush())
+                                        .map_err(|error| RuntimeError::new(error.to_string()))?;
+                                }
+                            }
+                        }
+                        ContentBlockDelta::SignatureDelta { .. } => {}
                    },
                    ApiStreamEvent::ContentBlockStop(_) => {
                        if let Some(rendered) = markdown_stream.flush(&renderer) {
@@ -3237,7 +3241,8 @@ fn final_assistant_text(summary: &runtime::TurnSummary) -> String {
                .iter()
                .filter_map(|block| match block {
                    ContentBlock::Text { text } => Some(text.as_str()),
-                    _ => None,
+                    ContentBlock::ToolUse { .. }
+                    | ContentBlock::ToolResult { .. } => None,
                })
                .collect::<Vec<_>>()
                .join("")
@@ -3256,7 +3261,8 @@ fn collect_tool_uses(summary: &runtime::TurnSummary) -> Vec<serde_json::Value> {
                "name": name,
                "input": input,
            })),
-            _ => None,
+            ContentBlock::Text { .. }
+            | ContentBlock::ToolResult { .. } => None,
        })
        .collect()
 }
@@ -3278,7 +3284,8 @@ fn collect_tool_results(summary: &runtime::TurnSummary) -> Vec<serde_json::Value
                "output": output,
                "is_error": is_error,
            })),
-            _ => None,
+            ContentBlock::Text { .. }
+            | ContentBlock::ToolUse { .. } => None,
        })
        .collect()
 }
@@ -3819,7 +3826,15 @@ fn push_output_block(
            };
            *pending_tool = Some((id, name, initial_input));
        }
-        OutputContentBlock::Thinking { .. } | OutputContentBlock::RedactedThinking { .. } => {}
+        OutputContentBlock::Thinking { thinking, .. } => {
+            if !thinking.is_empty() {
+                // Dimmed style for thinking
+                write!(out, "\x1b[2m{thinking}\x1b[0m")
+                    .and_then(|()| out.flush())
+                    .map_err(|error| RuntimeError::new(error.to_string()))?;
+            }
+        }
+        OutputContentBlock::RedactedThinking { .. } => {}
    }
    Ok(())
 }
@@ -3907,7 +3922,7 @@ impl ToolExecutor for CliToolExecutor {
 }

 fn permission_policy(mode: PermissionMode, tool_registry: &GlobalToolRegistry) -> PermissionPolicy {
-    tool_registry.permission_specs(None).into_iter().fold(
+    tool_registry.permission_specs(None).unwrap_or_default().into_iter().fold(
        PermissionPolicy::new(mode),
        |policy, (name, required_permission)| {
            policy.with_tool_requirement(name, required_permission)
@@ -4690,39 +4705,39 @@ mod tests {
    #[test]
    fn clear_command_requires_explicit_confirmation_flag() {
        assert_eq!(
-            SlashCommand::parse("/clear"),
-            Some(SlashCommand::Clear { confirm: false })
+            SlashCommand::parse("/clear").map_err(|e| e.to_string()),
+            Ok(Some(SlashCommand::Clear { confirm: false }))
        );
        assert_eq!(
-            SlashCommand::parse("/clear --confirm"),
-            Some(SlashCommand::Clear { confirm: true })
+            SlashCommand::parse("/clear --confirm").map_err(|e| e.to_string()),
+            Ok(Some(SlashCommand::Clear { confirm: true }))
        );
    }

    #[test]
    fn parses_resume_and_config_slash_commands() {
        assert_eq!(
-            SlashCommand::parse("/resume saved-session.json"),
-            Some(SlashCommand::Resume {
+            SlashCommand::parse("/resume saved-session.json").map_err(|e| e.to_string()),
+            Ok(Some(SlashCommand::Resume {
                session_path: Some("saved-session.json".to_string())
-            })
+            }))
        );
        assert_eq!(
-            SlashCommand::parse("/clear --confirm"),
-            Some(SlashCommand::Clear { confirm: true })
+            SlashCommand::parse("/clear --confirm").map_err(|e| e.to_string()),
+            Ok(Some(SlashCommand::Clear { confirm: true }))
        );
        assert_eq!(
-            SlashCommand::parse("/config"),
-            Some(SlashCommand::Config { section: None })
+            SlashCommand::parse("/config").map_err(|e| e.to_string()),
+            Ok(Some(SlashCommand::Config { section: None }))
        );
        assert_eq!(
-            SlashCommand::parse("/config env"),
-            Some(SlashCommand::Config {
+            SlashCommand::parse("/config env").map_err(|e| e.to_string()),
+            Ok(Some(SlashCommand::Config {
                section: Some("env".to_string())
-            })
+            }))
        );
-        assert_eq!(SlashCommand::parse("/memory"), Some(SlashCommand::Memory));
-        assert_eq!(SlashCommand::parse("/init"), Some(SlashCommand::Init));
+        assert_eq!(SlashCommand::parse("/memory").map_err(|e| e.to_string()), Ok(Some(SlashCommand::Memory)));
+        assert_eq!(SlashCommand::parse("/init").map_err(|e| e.to_string()), Ok(Some(SlashCommand::Init)));
    }

    #[test]
@@ -0,0 +1,55 @@
+# 命令模块 (commands)
+
+本模块负责定义和管理 Claw 交互界面中使用的“斜杠命令”(Slash Commands)，并提供相关的解析和执行逻辑。
+
+## 概览
+
+`commands` 模块的主要职责包括：
+- 定义所有可用的斜杠命令及其元数据（别名、说明、类别等）。
+- 提供命令注册表 (`CommandRegistry`)，用于在 CLI 中发现和分发命令。
+- 实现复杂的管理命令，如插件管理 (`/plugins`)、智能体查看 (`/agents`) 和技能查看 (`/skills`)。
+- 提供命令建议功能，支持基于编辑距离 (Levenshtein distance) 的模糊匹配。
+
+## 关键特性
+
+- **斜杠命令规范 (SlashCommandSpec)**：每个命令都包含详尽的元数据，包括所属类别（核心、工作区、会话、Git、自动化）以及是否支持在恢复会话时执行。
+- **命令分类**：
+  - **核心 (Core)**：`/help`, `/status`, `/model`, `/permissions`, `/cost` 等。
+  - **工作区 (Workspace)**：`/config`, `/memory`, `/diff`, `/teleport` 等。
+  - **会话 (Session)**：`/clear`, `/resume`, `/export`, `/session` 等。
+  - **Git 交互**：`/branch`, `/commit`, `/pr`, `/issue` 等。
+  - **自动化 (Automation)**：`/plugins`, `/agents`, `/skills`, `/ultraplan` 等。
+- **模糊匹配与建议**：当用户输入错误的命令时，系统会自动推荐最接近的合法命令。
+- **插件集成**：`/plugins` 命令允许用户动态安装、启用、禁用或卸载插件，并能通知运行时重新加载环境。
+
+## 实现逻辑
+
+### 核心模块
+
+- **`lib.rs`**: 包含了绝大部分逻辑。
+  - **`SlashCommand` 枚举**: 定义了所有命令的强类型表示。
+  - **`SlashCommandSpec` 结构体**: 存储命令的静态配置信息。
+  - **`handle_plugins_slash_command`**: 处理复杂的插件管理工作流。
+  - **`suggest_slash_commands`**: 实现基于 Levenshtein 距离的建议算法。
+
+### 工作流程
+
+1. 用户在 REPL 中输入以 `/` 开头的字符串。
+2. `claw-cli` 调用 `SlashCommand::parse` 进行解析。
+3. 解析后的命令被分发到相应的处理器。
+4. 处理结果（通常包含要显示给用户的消息，以及可选的会话更新或运行时重新加载请求）返回给 CLI。
+
+## 使用示例 (内部)
+
+```rust
+use commands::{SlashCommand, suggest_slash_commands};
+
+// 解析命令
+if let Some(cmd) = SlashCommand::parse("/model sonnet") {
+    // 处理模型切换逻辑
+}
+
+// 获取建议
+let suggestions = suggest_slash_commands("hpel", 3);
+// 返回 ["/help"]
+```
@@ -0,0 +1,48 @@
+# 兼容性测试套件模块 (compat-harness)
+
+本模块提供了一套工具，专门用于分析和提取上游引用实现（如原始的 `claude-code` TypeScript 源码）中的元数据，以确保 Rust 版本的实现与其保持功能兼容。
+
+## 概览
+
+`compat-harness` 的主要职责是：
+- 定位上游仓库的源码路径。
+- 从 TypeScript 源码文件中提取命令 (`commands`)、工具 (`tools`) 和启动阶段 (`bootstrap phases`) 的定义。
+- 自动生成功能清单 (`ExtractedManifest`)，供运行时或测试使用，以验证 Rust 版本的覆盖率。
+
+## 关键特性
+
+- **上游路径解析 (UpstreamPaths)**：能够自动识别多种常见的上游仓库目录结构，并支持通过环境变量 `CLAUDE_CODE_UPSTREAM` 进行覆盖。
+- **静态代码分析**：通过解析 TypeScript 源码，识别特定的代码模式（如 `export const INTERNAL_ONLY_COMMANDS` 或基于 `feature()` 的功能开关）。
+- **清单提取 (Manifest Extraction)**：
+  - **命令提取**：识别内置命令、仅限内部使用的命令以及受功能开关控制的命令。
+  - **工具提取**：识别基础工具和条件加载的工具。
+  - **启动计划提取**：分析 CLI 入口文件，重建启动时的各个阶段（如 `FastPathVersion`, `MainRuntime` 等）。
+
+## 实现逻辑
+
+### 核心模块
+
+- **`lib.rs`**: 包含了核心的提取逻辑。
+  - **`UpstreamPaths` 结构体**: 封装了寻找 `commands.ts`、`tools.ts` 和 `cli.tsx` 的逻辑。
+  - **`extract_commands` & `extract_tools`**: 使用字符串解析技术，识别 TypeScript 的 `import` 和赋值操作，提取符号名称。
+  - **`extract_bootstrap_plan`**: 搜索特定的标志性字符串（如 `--version` 或 `daemon-worker`），从而推断出上游程序的启动流程。
+
+### 工作流程
+
+1. 模块根据预设路径或环境变量寻找上游 `claude-code` 仓库。
+2. 读取关键的 `.ts` 或 `.tsx` 文件内容。
+3. 执行正则表达式风格的行解析，提取出所有定义的命令和工具名称。
+4. 将提取结果组织成 `ExtractedManifest` 对象。
+
+## 使用示例 (内部测试)
+
+```rust
+use compat_harness::{UpstreamPaths, extract_manifest};
+
+// 指定工作区目录，自动寻找上游路径
+let paths = UpstreamPaths::from_workspace_dir("path/to/workspace");
+// 提取功能清单
+if let Ok(manifest) = extract_manifest(&paths) {
+    println!("上游发现 {} 个工具", manifest.tools.entries().len());
+}
+```
@@ -70,16 +70,12 @@ fn upstream_repo_candidates(primary_repo_root: &Path) -> Vec<PathBuf> {
    }

    for ancestor in primary_repo_root.ancestors().take(4) {
-        candidates.push(ancestor.join("claude-code"));
+        candidates.push(ancestor.join("claw-code"));
        candidates.push(ancestor.join("clawd-code"));
    }

-    candidates.push(
-        primary_repo_root
-            .join("reference-source")
-            .join("claude-code"),
-    );
-    candidates.push(primary_repo_root.join("vendor").join("claude-code"));
+    candidates.push(primary_repo_root.join("reference-source").join("claw-code"));
+    candidates.push(primary_repo_root.join("vendor").join("claw-code"));

    let mut deduped = Vec::new();
    for candidate in candidates {
@@ -0,0 +1,56 @@
+# LSP 模块 (lsp)
+
+本模块实现了语言服务协议 (Language Server Protocol, LSP) 的客户端功能，允许系统通过集成的编程语言服务器获取代码的语义信息、错误诊断和符号导航。
+
+## 概览
+
+`lsp` 模块的主要职责是：
+- 管理多个 LSP 服务器的生命周期（启动、初始化、关闭）。
+- 与服务器进行异步 JSON-RPC 通信。
+- 提供跨语言的代码智能功能，如：
+  - **转到定义 (Go to Definition)**
+  - **查找引用 (Find References)**
+  - **工作区诊断 (Workspace Diagnostics)**
+- 为 AI 提示词 (Prompt) 提供上下文增强，将代码中的实时错误和符号关系反馈给 LLM。
+
+## 关键特性
+
+- **LspManager**: 核心管理类，负责协调不同语言的服务器配置和文档状态。
+- **上下文增强 (Context Enrichment)**：定义了 `LspContextEnrichment` 结构，能够将复杂的 LSP 响应（如诊断信息和定义）转换为易于 AI 理解的 Markdown 格式。
+- **多服务器支持**：支持根据文件扩展名将请求路由到不同的语言服务器（如 `rust-analyzer`, `pyright` 等）。
+- **同步机制**：处理文档的 `didOpen`、`didChange` 和 `didSave` 消息，确保服务器拥有最新的代码视图。
+
+## 实现逻辑
+
+### 核心模块
+
+- **`manager.rs`**: 实现了 `LspManager`。它维护一个服务器池，并提供高层 API 来执行跨服务器的请求。
+- **`client.rs`**: 实现底层的 LSP 客户端逻辑，处理基于 `tokio` 的异步 I/O 和 JSON-RPC 消息的分帧与解析。
+- **`types.rs`**: 定义了本模块使用的专用数据类型，并对 `lsp-types` 库中的类型进行了简化和包装，以便于内部使用。
+- **`error.rs`**: 定义了 LSP 相关的错误处理。
+
+### 工作流程
+
+1. 系统根据配置初始化 `LspManager`。
+2. 当打开一个文件时，`LspManager` 启动相应的服务器并发送 `initialize` 请求。
+3. `LspManager` 跟踪文档的打开状态，并在内容变化时同步到服务器。
+4. 当需要对某个符号进行分析时，调用 `go_to_definition` 等方法，模块负责发送请求并解析返回的 `Location`。
+5. 诊断信息异步通过 `textDocument/publishDiagnostics` 通知到达，模块会缓存这些信息供后续查询。
+
+## 使用示例 (内部)
+
+```rust
+use lsp::{LspManager, LspServerConfig};
+
+// 配置并初始化管理器
+let configs = vec![LspServerConfig {
+    name: "rust-analyzer".to_string(),
+    command: "rust-analyzer".to_string(),
+    ..Default::default()
+}];
+let manager = LspManager::new(configs)?;
+
+// 获取某个位置的上下文增强信息
+let enrichment = manager.context_enrichment(&file_path, position).await?;
+println!("{}", enrichment.render_prompt_section());
+```
@@ -15,11 +15,13 @@ use tokio::sync::{oneshot, Mutex};
 use crate::error::LspError;
 use crate::types::{LspServerConfig, SymbolLocation};

+type PendingRequestMap = BTreeMap<i64, oneshot::Sender<Result<Value, LspError>>>;
+
 pub(crate) struct LspClient {
    config: LspServerConfig,
    writer: Mutex<BufWriter<ChildStdin>>,
    child: Mutex<Child>,
-    pending_requests: Arc<Mutex<BTreeMap<i64, oneshot::Sender<Result<Value, LspError>>>>>,
+    pending_requests: Arc<Mutex<PendingRequestMap>>,
    diagnostics: Arc<Mutex<BTreeMap<String, Vec<Diagnostic>>>>,
    open_documents: Mutex<BTreeMap<PathBuf, i32>>,
    next_request_id: AtomicI64,
@@ -59,7 +61,7 @@ impl LspClient {

        client.spawn_reader(stdout);
        if let Some(stderr) = stderr {
-            client.spawn_stderr_drain(stderr);
+            Self::spawn_stderr_drain(stderr);
        }
        client.initialize().await?;
        Ok(client)
@@ -282,8 +284,8 @@ impl LspClient {
            if let Err(error) = result {
                let mut pending = pending_requests.lock().await;
                let drained = pending
-                    .iter()
-                    .map(|(id, _)| *id)
+                    .keys()
+                    .copied()
                    .collect::<Vec<_>>();
                for id in drained {
                    if let Some(sender) = pending.remove(&id) {
@@ -294,7 +296,7 @@ impl LspClient {
        });
    }

-    fn spawn_stderr_drain<R>(&self, stderr: R)
+    fn spawn_stderr_drain<R>(stderr: R)
    where
        R: AsyncRead + Unpin + Send + 'static,
    {
@@ -41,6 +41,7 @@ mod tests {
        })
    }

+    #[allow(clippy::too_many_lines)]
    fn write_mock_server_script(root: &std::path::Path) -> PathBuf {
        let script_path = root.join("mock_lsp_server.py");
        fs::write(
@@ -0,0 +1,18 @@
+[package]
+name = "mock-anthropic-service"
+version.workspace = true
+edition.workspace = true
+license.workspace = true
+publish.workspace = true
+
+[[bin]]
+name = "mock-anthropic-service"
+path = "src/main.rs"
+
+[dependencies]
+api = { path = "../api" }
+serde_json.workspace = true
+tokio = { version = "1", features = ["io-util", "macros", "net", "rt-multi-thread", "signal", "sync"] }
+
+[lints]
+workspace = true
@@ -0,0 +1,34 @@
+use std::env;
+
+use mock_anthropic_service::MockAnthropicService;
+
+#[tokio::main(flavor = "multi_thread")]
+async fn main() -> Result<(), Box<dyn std::error::Error>> {
+    let mut bind_addr = String::from("127.0.0.1:0");
+    let mut args = env::args().skip(1);
+    while let Some(arg) = args.next() {
+        match arg.as_str() {
+            "--bind" => {
+                bind_addr = args
+                    .next()
+                    .ok_or_else(|| "missing value for --bind".to_string())?;
+            }
+            flag if flag.starts_with("--bind=") => {
+                bind_addr = flag[7..].to_string();
+            }
+            "--help" | "-h" => {
+                println!("Usage: mock-anthropic-service [--bind HOST:PORT]");
+                return Ok(());
+            }
+            other => {
+                return Err(format!("unsupported argument: {other}").into());
+            }
+        }
+    }
+
+    let server = MockAnthropicService::spawn_on(&bind_addr).await?;
+    println!("MOCK_ANTHROPIC_BASE_URL={}", server.base_url());
+    tokio::signal::ctrl_c().await?;
+    drop(server);
+    Ok(())
+}
@@ -0,0 +1,59 @@
+# 插件模块 (plugins)
+
+本模块实现了 Claw 的插件系统，允许通过外部扩展来增强 AI 的功能，包括自定义工具、命令以及在工具执行前后运行的钩子 (Hooks)。
+
+## 概览
+
+`plugins` 模块的主要职责是：
+- 定义插件的清单格式 (`plugin.json`) 和元数据结构。
+- 管理插件的完整生命周期：安装、加载、初始化、启用/禁用、更新和卸载。
+- 提供插件类型的抽象：
+  - **Builtin (内置)**：编译在程序内部的插件。
+  - **Bundled (绑定)**：随应用程序分发但作为独立文件存在的插件。
+  - **External (外部)**：用户自行安装或从远程仓库下载的插件。
+- 实现插件隔离与执行机制，支持插件定义的自定义工具。
+
+## 关键特性
+
+- **插件清单 (PluginManifest)**：每个插件必须包含一个 `plugin.json`，详细说明其名称、版本、所需权限、钩子、生命周期脚本以及它所暴露的工具。
+- **自定义工具 (PluginTool)**：插件可以定义全新的工具供 AI 调用。这些工具在执行时被作为独立的外部进程启动。
+- **钩子系统 (Hooks)**：支持 `PreToolUse` 和 `PostToolUse` 钩子，允许插件在 AI 调用任何工具之前或之后执行特定的逻辑。
+- **生命周期管理**：提供 `Init` 和 `Shutdown` 阶段，允许插件在加载时进行环境准备，在卸载或关闭时进行清理。
+- **权限模型**：强制要求插件声明权限（`read`, `write`, `execute`），并为定义的工具指定安全级别（`read-only`, `workspace-write`, `danger-full-access`）。
+
+## 实现逻辑
+
+### 核心模块
+
+- **`lib.rs`**: 包含了插件定义的各种结构体（Manifest, Metadata, Tool, Permission 等）以及插件特性的定义。
+- **`manager.rs`**: 实现了 `PluginManager`，负责插件在磁盘上的组织、注册表的维护以及安装/更新逻辑。
+- **`hooks.rs`**: 实现了钩子执行器 (`HookRunner`)，负责在正确的时机触发插件定义的脚本。
+
+### 插件加载与执行流程
+
+1. `PluginManager` 扫描指定的目录（内置、绑定及外部安装目录）。
+2. 读取并验证每个插件的 `plugin.json`。
+3. 如果插件被启用，则初始化该插件并将其定义的工具注册到系统的全局工具注册表中。
+4. 当 AI 调用插件工具时，系统根据清单中定义的命令行信息启动一个子进程，并通过标准输入/环境变量传递参数。
+5. 钩子逻辑会在工具执行的生命周期内被自动触发。
+
+## 使用示例 (插件定义样例)
+
+```json
+{
+  "name": "my-custom-plugin",
+  "version": "1.0.0",
+  "description": "一个演示插件",
+  "permissions": ["read", "execute"],
+  "tools": [
+    {
+      "name": "custom_search",
+      "description": "执行自定义搜索",
+      "inputSchema": { "type": "object", "properties": { "query": { "type": "string" } } },
+      "command": "python3",
+      "args": ["search_script.py"],
+      "requiredPermission": "read-only"
+    }
+  ]
+}
+```
@@ -1,5 +1,4 @@
 use std::ffi::OsStr;
-use std::path::Path;
 use std::process::Command;

 use serde_json::json;
@@ -10,6 +9,7 @@ use crate::{PluginError, PluginHooks, PluginRegistry};
 pub enum HookEvent {
    PreToolUse,
    PostToolUse,
+    PostToolUseFailure,
 }

 impl HookEvent {
@@ -17,6 +17,7 @@ impl HookEvent {
        match self {
            Self::PreToolUse => "PreToolUse",
            Self::PostToolUse => "PostToolUse",
+            Self::PostToolUseFailure => "PostToolUseFailure",
        }
    }
 }
@@ -24,6 +25,7 @@ impl HookEvent {
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct HookRunResult {
    denied: bool,
+    failed: bool,
    messages: Vec<String>,
 }

@@ -32,6 +34,7 @@ impl HookRunResult {
    pub fn allow(messages: Vec<String>) -> Self {
        Self {
            denied: false,
+            failed: false,
            messages,
        }
    }
@@ -41,6 +44,11 @@ impl HookRunResult {
        self.denied
    }

+    #[must_use]
+    pub fn is_failed(&self) -> bool {
+        self.failed
+    }
+
    #[must_use]
    pub fn messages(&self) -> &[String] {
        &self.messages
@@ -64,7 +72,7 @@ impl HookRunner {

    #[must_use]
    pub fn run_pre_tool_use(&self, tool_name: &str, tool_input: &str) -> HookRunResult {
-        self.run_commands(
+        Self::run_commands(
            HookEvent::PreToolUse,
            &self.hooks.pre_tool_use,
            tool_name,
@@ -82,7 +90,7 @@ impl HookRunner {
        tool_output: &str,
        is_error: bool,
    ) -> HookRunResult {
-        self.run_commands(
+        Self::run_commands(
            HookEvent::PostToolUse,
            &self.hooks.post_tool_use,
            tool_name,
@@ -92,8 +100,24 @@ impl HookRunner {
        )
    }

-    fn run_commands(
+    #[must_use]
+    pub fn run_post_tool_use_failure(
        &self,
+        tool_name: &str,
+        tool_input: &str,
+        tool_error: &str,
+    ) -> HookRunResult {
+        Self::run_commands(
+            HookEvent::PostToolUseFailure,
+            &self.hooks.post_tool_use_failure,
+            tool_name,
+            tool_input,
+            Some(tool_error),
+            true,
+        )
+    }
+
+    fn run_commands(
        event: HookEvent,
        commands: &[String],
        tool_name: &str,
@@ -105,20 +129,12 @@ impl HookRunner {
            return HookRunResult::allow(Vec::new());
        }

-        let payload = json!({
-            "hook_event_name": event.as_str(),
-            "tool_name": tool_name,
-            "tool_input": parse_tool_input(tool_input),
-            "tool_input_json": tool_input,
-            "tool_output": tool_output,
-            "tool_result_is_error": is_error,
-        })
-        .to_string();
+        let payload = hook_payload(event, tool_name, tool_input, tool_output, is_error).to_string();

        let mut messages = Vec::new();

        for command in commands {
-            match self.run_command(
+            match Self::run_command(
                command,
                event,
                tool_name,
@@ -138,19 +154,26 @@ impl HookRunner {
                    }));
                    return HookRunResult {
                        denied: true,
+                        failed: false,
+                        messages,
+                    };
+                }
+                HookCommandOutcome::Failed { message } => {
+                    messages.push(message);
+                    return HookRunResult {
+                        denied: false,
+                        failed: true,
                        messages,
                    };
                }
-                HookCommandOutcome::Warn { message } => messages.push(message),
            }
        }

        HookRunResult::allow(messages)
    }

-    #[allow(clippy::too_many_arguments, clippy::unused_self)]
+    #[allow(clippy::too_many_arguments)]
    fn run_command(
-        &self,
        command: &str,
        event: HookEvent,
        tool_name: &str,
@@ -179,7 +202,7 @@ impl HookRunner {
                match output.status.code() {
                    Some(0) => HookCommandOutcome::Allow { message },
                    Some(2) => HookCommandOutcome::Deny { message },
-                    Some(code) => HookCommandOutcome::Warn {
+                    Some(code) => HookCommandOutcome::Failed {
                        message: format_hook_warning(
                            command,
                            code,
@@ -187,7 +210,7 @@ impl HookRunner {
                            stderr.as_str(),
                        ),
                    },
-                    None => HookCommandOutcome::Warn {
+                    None => HookCommandOutcome::Failed {
                        message: format!(
                            "{} hook `{command}` terminated by signal while handling `{tool_name}`",
                            event.as_str()
@@ -195,7 +218,7 @@ impl HookRunner {
                    },
                }
            }
-            Err(error) => HookCommandOutcome::Warn {
+            Err(error) => HookCommandOutcome::Failed {
                message: format!(
                    "{} hook `{command}` failed to start for `{tool_name}`: {error}",
                    event.as_str()
@@ -208,7 +231,34 @@ impl HookRunner {
 enum HookCommandOutcome {
    Allow { message: Option<String> },
    Deny { message: Option<String> },
-    Warn { message: String },
+    Failed { message: String },
+}
+
+fn hook_payload(
+    event: HookEvent,
+    tool_name: &str,
+    tool_input: &str,
+    tool_output: Option<&str>,
+    is_error: bool,
+) -> serde_json::Value {
+    match event {
+        HookEvent::PostToolUseFailure => json!({
+            "hook_event_name": event.as_str(),
+            "tool_name": tool_name,
+            "tool_input": parse_tool_input(tool_input),
+            "tool_input_json": tool_input,
+            "tool_error": tool_output,
+            "tool_result_is_error": true,
+        }),
+        _ => json!({
+            "hook_event_name": event.as_str(),
+            "tool_name": tool_name,
+            "tool_input": parse_tool_input(tool_input),
+            "tool_input_json": tool_input,
+            "tool_output": tool_output,
+            "tool_result_is_error": is_error,
+        }),
+    }
 }

 fn parse_tool_input(tool_input: &str) -> serde_json::Value {
@@ -216,8 +266,7 @@ fn parse_tool_input(tool_input: &str) -> serde_json::Value {
 }

 fn format_hook_warning(command: &str, code: i32, stdout: Option<&str>, stderr: &str) -> String {
-    let mut message =
-        format!("Hook `{command}` exited with status {code}; allowing tool execution to continue");
+    let mut message = format!("Hook `{command}` exited with status {code}");
    if let Some(stdout) = stdout.filter(|stdout| !stdout.is_empty()) {
        message.push_str(": ");
        message.push_str(stdout);
@@ -287,7 +336,28 @@ impl CommandWithStdin {
        let mut child = self.command.spawn()?;
        if let Some(mut child_stdin) = child.stdin.take() {
            use std::io::Write as _;
-            child_stdin.write_all(stdin)?;
+            // Tolerate BrokenPipe: a hook script that runs to completion
+            // (or exits early without reading stdin) closes its stdin
+            // before the parent finishes writing the JSON payload, and
+            // the kernel raises EPIPE on the parent's write_all. That is
+            // not a hook failure — the child still exited cleanly and we
+            // still need to wait_with_output() to capture stdout/stderr
+            // and the real exit code. Other write errors (e.g. EIO,
+            // permission, OOM) still propagate.
+            //
+            // This was the root cause of the Linux CI flake on
+            // hooks::tests::collects_and_runs_hooks_from_enabled_plugins
+            // (ROADMAP #25, runs 24120271422 / 24120538408 / 24121392171
+            // / 24121776826): the test hook scripts run in microseconds
+            // and the parent's stdin write races against child exit.
+            // macOS pipes happen to buffer the small payload before the
+            // child exits; Linux pipes do not, so the race shows up
+            // deterministically on ubuntu runners.
+            match child_stdin.write_all(stdin) {
+                Ok(()) => {}
+                Err(error) if error.kind() == std::io::ErrorKind::BrokenPipe => {}
+                Err(error) => return Err(error),
+            }
        }
        child.wait_with_output()
    }
@@ -309,23 +379,55 @@ mod tests {
        std::env::temp_dir().join(format!("plugins-hook-runner-{label}-{nanos}"))
    }

-    fn write_hook_plugin(root: &Path, name: &str, pre_message: &str, post_message: &str) {
-        fs::create_dir_all(root.join(".claw-plugin")).expect("manifest dir");
+    fn make_executable(path: &Path) {
+        #[cfg(unix)]
+        {
+            use std::os::unix::fs::PermissionsExt;
+            let perms = fs::Permissions::from_mode(0o755);
+            fs::set_permissions(path, perms)
+                .unwrap_or_else(|e| panic!("chmod +x {}: {e}", path.display()));
+        }
+        #[cfg(not(unix))]
+        let _ = path;
+    }
+
+    fn write_hook_plugin(
+        root: &Path,
+        name: &str,
+        pre_message: &str,
+        post_message: &str,
+        failure_message: &str,
+    ) {
+        fs::create_dir_all(root.join(".claude-plugin")).expect("manifest dir");
        fs::create_dir_all(root.join("hooks")).expect("hooks dir");
+
+        let pre_path = root.join("hooks").join("pre.sh");
        fs::write(
-            root.join("hooks").join("pre.sh"),
+            &pre_path,
            format!("#!/bin/sh\nprintf '%s\\n' '{pre_message}'\n"),
        )
        .expect("write pre hook");
+        make_executable(&pre_path);
+
+        let post_path = root.join("hooks").join("post.sh");
        fs::write(
-            root.join("hooks").join("post.sh"),
+            &post_path,
            format!("#!/bin/sh\nprintf '%s\\n' '{post_message}'\n"),
        )
        .expect("write post hook");
+        make_executable(&post_path);
+
+        let failure_path = root.join("hooks").join("failure.sh");
        fs::write(
-            root.join(".claw-plugin").join("plugin.json"),
+            &failure_path,
+            format!("#!/bin/sh\nprintf '%s\\n' '{failure_message}'\n"),
+        )
+        .expect("write failure hook");
+        make_executable(&failure_path);
+        fs::write(
+            root.join(".claude-plugin").join("plugin.json"),
            format!(
-                "{{\n  \"name\": \"{name}\",\n  \"version\": \"1.0.0\",\n  \"description\": \"hook plugin\",\n  \"hooks\": {{\n    \"PreToolUse\": [\"./hooks/pre.sh\"],\n    \"PostToolUse\": [\"./hooks/post.sh\"]\n  }}\n}}"
+                "{{\n  \"name\": \"{name}\",\n  \"version\": \"1.0.0\",\n  \"description\": \"hook plugin\",\n  \"hooks\": {{\n    \"PreToolUse\": [\"./hooks/pre.sh\"],\n    \"PostToolUse\": [\"./hooks/post.sh\"],\n    \"PostToolUseFailure\": [\"./hooks/failure.sh\"]\n  }}\n}}"
            ),
        )
        .expect("write plugin manifest");
@@ -333,6 +435,7 @@ mod tests {

    #[test]
    fn collects_and_runs_hooks_from_enabled_plugins() {
+        // given
        let config_home = temp_dir("config");
        let first_source_root = temp_dir("source-a");
        let second_source_root = temp_dir("source-b");
@@ -341,12 +444,14 @@ mod tests {
            "first",
            "plugin pre one",
            "plugin post one",
+            "plugin failure one",
        );
        write_hook_plugin(
            &second_source_root,
            "second",
            "plugin pre two",
            "plugin post two",
+            "plugin failure two",
        );

        let mut manager = PluginManager::new(PluginManagerConfig::new(&config_home));
@@ -358,8 +463,10 @@ mod tests {
            .expect("second plugin install should succeed");
        let registry = manager.plugin_registry().expect("registry should build");

+        // when
        let runner = HookRunner::from_registry(&registry).expect("plugin hooks should load");

+        // then
        assert_eq!(
            runner.run_pre_tool_use("Read", r#"{"path":"README.md"}"#),
            HookRunResult::allow(vec![
@@ -374,6 +481,13 @@ mod tests {
                "plugin post two".to_string(),
            ])
        );
+        assert_eq!(
+            runner.run_post_tool_use_failure("Read", r#"{"path":"README.md"}"#, "tool failed",),
+            HookRunResult::allow(vec![
+                "plugin failure one".to_string(),
+                "plugin failure two".to_string(),
+            ])
+        );

        let _ = fs::remove_dir_all(config_home);
        let _ = fs::remove_dir_all(first_source_root);
@@ -382,14 +496,68 @@ mod tests {

    #[test]
    fn pre_tool_use_denies_when_plugin_hook_exits_two() {
+        // given
        let runner = HookRunner::new(crate::PluginHooks {
            pre_tool_use: vec!["printf 'blocked by plugin'; exit 2".to_string()],
            post_tool_use: Vec::new(),
+            post_tool_use_failure: Vec::new(),
        });

+        // when
        let result = runner.run_pre_tool_use("Bash", r#"{"command":"pwd"}"#);

+        // then
        assert!(result.is_denied());
        assert_eq!(result.messages(), &["blocked by plugin".to_string()]);
    }
+
+    #[test]
+    fn propagates_plugin_hook_failures() {
+        // given
+        let runner = HookRunner::new(crate::PluginHooks {
+            pre_tool_use: vec![
+                "printf 'broken plugin hook'; exit 1".to_string(),
+                "printf 'later plugin hook'".to_string(),
+            ],
+            post_tool_use: Vec::new(),
+            post_tool_use_failure: Vec::new(),
+        });
+
+        // when
+        let result = runner.run_pre_tool_use("Bash", r#"{"command":"pwd"}"#);
+
+        // then
+        assert!(result.is_failed());
+        assert!(result
+            .messages()
+            .iter()
+            .any(|message| message.contains("broken plugin hook")));
+        assert!(!result
+            .messages()
+            .iter()
+            .any(|message| message == "later plugin hook"));
+    }
+
+    #[test]
+    #[cfg(unix)]
+    fn generated_hook_scripts_are_executable() {
+        use std::os::unix::fs::PermissionsExt;
+
+        // given
+        let root = temp_dir("exec-guard");
+        write_hook_plugin(&root, "exec-check", "pre", "post", "fail");
+
+        // then
+        for script in ["pre.sh", "post.sh", "failure.sh"] {
+            let path = root.join("hooks").join(script);
+            let mode = fs::metadata(&path)
+                .unwrap_or_else(|e| panic!("{script} metadata: {e}"))
+                .permissions()
+                .mode();
+            assert!(
+                mode & 0o111 != 0,
+                "{script} must have at least one execute bit set, got mode {mode:#o}"
+            );
+        }
+    }
 }
@@ -18,7 +18,7 @@ const BUNDLED_MARKETPLACE: &str = "bundled";
 const SETTINGS_FILE_NAME: &str = "settings.json";
 const REGISTRY_FILE_NAME: &str = "installed.json";
 const MANIFEST_FILE_NAME: &str = "plugin.json";
-const MANIFEST_RELATIVE_PATH: &str = ".claw-plugin/plugin.json";
+const MANIFEST_RELATIVE_PATH: &str = ".claude-plugin/plugin.json";

 #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
 #[serde(rename_all = "lowercase")]
@@ -67,12 +67,16 @@ pub struct PluginHooks {
    pub pre_tool_use: Vec<String>,
    #[serde(rename = "PostToolUse", default)]
    pub post_tool_use: Vec<String>,
+    #[serde(rename = "PostToolUseFailure", default)]
+    pub post_tool_use_failure: Vec<String>,
 }

 impl PluginHooks {
    #[must_use]
    pub fn is_empty(&self) -> bool {
-        self.pre_tool_use.is_empty() && self.post_tool_use.is_empty()
+        self.pre_tool_use.is_empty()
+            && self.post_tool_use.is_empty()
+            && self.post_tool_use_failure.is_empty()
    }

    #[must_use]
@@ -85,6 +89,9 @@ impl PluginHooks {
            .post_tool_use
            .extend(other.post_tool_use.iter().cloned());
        merged
+            .post_tool_use_failure
+            .extend(other.post_tool_use_failure.iter().cloned());
+        merged
    }
 }

@@ -302,14 +309,14 @@ impl PluginTool {
            .stdin(Stdio::piped())
            .stdout(Stdio::piped())
            .stderr(Stdio::piped())
-            .env("CLAW_PLUGIN_ID", &self.plugin_id)
-            .env("CLAW_PLUGIN_NAME", &self.plugin_name)
-            .env("CLAW_TOOL_NAME", &self.definition.name)
-            .env("CLAW_TOOL_INPUT", &input_json);
+            .env("CLAWD_PLUGIN_ID", &self.plugin_id)
+            .env("CLAWD_PLUGIN_NAME", &self.plugin_name)
+            .env("CLAWD_TOOL_NAME", &self.definition.name)
+            .env("CLAWD_TOOL_INPUT", &input_json);
        if let Some(root) = &self.root {
            process
                .current_dir(root)
-                .env("CLAW_PLUGIN_ROOT", root.display().to_string());
+                .env("CLAWD_PLUGIN_ROOT", root.display().to_string());
        }

        let mut child = process.spawn()?;
@@ -648,6 +655,106 @@ pub struct PluginSummary {
    pub enabled: bool,
 }

+#[derive(Debug)]
+pub struct PluginLoadFailure {
+    pub plugin_root: PathBuf,
+    pub kind: PluginKind,
+    pub source: String,
+    error: Box<PluginError>,
+}
+
+impl PluginLoadFailure {
+    #[must_use]
+    pub fn new(plugin_root: PathBuf, kind: PluginKind, source: String, error: PluginError) -> Self {
+        Self {
+            plugin_root,
+            kind,
+            source,
+            error: Box::new(error),
+        }
+    }
+
+    #[must_use]
+    pub fn error(&self) -> &PluginError {
+        self.error.as_ref()
+    }
+}
+
+impl Display for PluginLoadFailure {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        write!(
+            f,
+            "failed to load {} plugin from `{}` (source: {}): {}",
+            self.kind,
+            self.plugin_root.display(),
+            self.source,
+            self.error()
+        )
+    }
+}
+
+#[derive(Debug)]
+pub struct PluginRegistryReport {
+    registry: PluginRegistry,
+    failures: Vec<PluginLoadFailure>,
+}
+
+impl PluginRegistryReport {
+    #[must_use]
+    pub fn new(registry: PluginRegistry, failures: Vec<PluginLoadFailure>) -> Self {
+        Self { registry, failures }
+    }
+
+    #[must_use]
+    pub fn registry(&self) -> &PluginRegistry {
+        &self.registry
+    }
+
+    #[must_use]
+    pub fn failures(&self) -> &[PluginLoadFailure] {
+        &self.failures
+    }
+
+    #[must_use]
+    pub fn has_failures(&self) -> bool {
+        !self.failures.is_empty()
+    }
+
+    #[must_use]
+    pub fn summaries(&self) -> Vec<PluginSummary> {
+        self.registry.summaries()
+    }
+
+    pub fn into_registry(self) -> Result<PluginRegistry, PluginError> {
+        if self.failures.is_empty() {
+            Ok(self.registry)
+        } else {
+            Err(PluginError::LoadFailures(self.failures))
+        }
+    }
+}
+
+#[derive(Debug, Default)]
+struct PluginDiscovery {
+    plugins: Vec<PluginDefinition>,
+    failures: Vec<PluginLoadFailure>,
+}
+
+impl PluginDiscovery {
+    fn push_plugin(&mut self, plugin: PluginDefinition) {
+        self.plugins.push(plugin);
+    }
+
+    fn push_failure(&mut self, failure: PluginLoadFailure) {
+        self.failures.push(failure);
+    }
+
+    fn extend(&mut self, other: Self) {
+        self.plugins.extend(other.plugins);
+        self.failures.extend(other.failures);
+    }
+}
+
 #[derive(Debug, Clone, Default, PartialEq)]
 pub struct PluginRegistry {
    plugins: Vec<RegisteredPlugin>,
@@ -802,6 +909,10 @@ pub enum PluginManifestValidationError {
        kind: &'static str,
        path: PathBuf,
    },
+    PathIsDirectory {
+        kind: &'static str,
+        path: PathBuf,
+    },
    InvalidToolInputSchema {
        tool_name: String,
    },
@@ -809,6 +920,9 @@ pub enum PluginManifestValidationError {
        tool_name: String,
        permission: String,
    },
+    UnsupportedManifestContract {
+        detail: String,
+    },
 }

 impl Display for PluginManifestValidationError {
@@ -838,6 +952,9 @@ impl Display for PluginManifestValidationError {
            Self::MissingPath { kind, path } => {
                write!(f, "{kind} path `{}` does not exist", path.display())
            }
+            Self::PathIsDirectory { kind, path } => {
+                write!(f, "{kind} path `{}` must point to a file", path.display())
+            }
            Self::InvalidToolInputSchema { tool_name } => {
                write!(
                    f,
@@ -851,6 +968,7 @@ impl Display for PluginManifestValidationError {
                f,
                "plugin tool `{tool_name}` requiredPermission `{permission}` must be read-only, workspace-write, or danger-full-access"
            ),
+            Self::UnsupportedManifestContract { detail } => f.write_str(detail),
        }
    }
 }
@@ -860,6 +978,7 @@ pub enum PluginError {
    Io(std::io::Error),
    Json(serde_json::Error),
    ManifestValidation(Vec<PluginManifestValidationError>),
+    LoadFailures(Vec<PluginLoadFailure>),
    InvalidManifest(String),
    NotFound(String),
    CommandFailed(String),
@@ -879,6 +998,15 @@ impl Display for PluginError {
                }
                Ok(())
            }
+            Self::LoadFailures(failures) => {
+                for (index, failure) in failures.iter().enumerate() {
+                    if index > 0 {
+                        write!(f, "; ")?;
+                    }
+                    write!(f, "{failure}")?;
+                }
+                Ok(())
+            }
            Self::InvalidManifest(message)
            | Self::NotFound(message)
            | Self::CommandFailed(message) => write!(f, "{message}"),
@@ -935,15 +1063,23 @@ impl PluginManager {
    }

    pub fn plugin_registry(&self) -> Result<PluginRegistry, PluginError> {
-        Ok(PluginRegistry::new(
-            self.discover_plugins()?
-                .into_iter()
-                .map(|plugin| {
-                    let enabled = self.is_enabled(plugin.metadata());
-                    RegisteredPlugin::new(plugin, enabled)
-                })
-                .collect(),
-        ))
+        self.plugin_registry_report()?.into_registry()
+    }
+
+    pub fn plugin_registry_report(&self) -> Result<PluginRegistryReport, PluginError> {
+        self.sync_bundled_plugins()?;
+
+        let mut discovery = PluginDiscovery::default();
+        discovery.plugins.extend(builtin_plugins());
+
+        let installed = self.discover_installed_plugins_with_failures()?;
+        discovery.extend(installed);
+
+        let external =
+            self.discover_external_directory_plugins_with_failures(&discovery.plugins)?;
+        discovery.extend(external);
+
+        Ok(self.build_registry_report(discovery))
    }

    pub fn list_plugins(&self) -> Result<Vec<PluginSummary>, PluginError> {
@@ -955,11 +1091,12 @@ impl PluginManager {
    }

    pub fn discover_plugins(&self) -> Result<Vec<PluginDefinition>, PluginError> {
-        self.sync_bundled_plugins()?;
-        let mut plugins = builtin_plugins();
-        plugins.extend(self.discover_installed_plugins()?);
-        plugins.extend(self.discover_external_directory_plugins(&plugins)?);
-        Ok(plugins)
+        Ok(self
+            .plugin_registry()?
+            .plugins
+            .into_iter()
+            .map(|plugin| plugin.definition)
+            .collect())
    }

    pub fn aggregated_hooks(&self) -> Result<PluginHooks, PluginError> {
@@ -1094,9 +1231,9 @@ impl PluginManager {
        })
    }

-    fn discover_installed_plugins(&self) -> Result<Vec<PluginDefinition>, PluginError> {
+    fn discover_installed_plugins_with_failures(&self) -> Result<PluginDiscovery, PluginError> {
        let mut registry = self.load_registry()?;
-        let mut plugins = Vec::new();
+        let mut discovery = PluginDiscovery::default();
        let mut seen_ids = BTreeSet::<String>::new();
        let mut seen_paths = BTreeSet::<PathBuf>::new();
        let mut stale_registry_ids = Vec::new();
@@ -1111,10 +1248,21 @@ impl PluginManager {
                || install_path.display().to_string(),
                |record| describe_install_source(&record.source),
            );
-            let plugin = load_plugin_definition(&install_path, kind, source, kind.marketplace())?;
-            if seen_ids.insert(plugin.metadata().id.clone()) {
-                seen_paths.insert(install_path);
-                plugins.push(plugin);
+            match load_plugin_definition(&install_path, kind, source.clone(), kind.marketplace()) {
+                Ok(plugin) => {
+                    if seen_ids.insert(plugin.metadata().id.clone()) {
+                        seen_paths.insert(install_path);
+                        discovery.push_plugin(plugin);
+                    }
+                }
+                Err(error) => {
+                    discovery.push_failure(PluginLoadFailure::new(
+                        install_path,
+                        kind,
+                        source,
+                        error,
+                    ));
+                }
            }
        }

@@ -1127,15 +1275,27 @@ impl PluginManager {
                stale_registry_ids.push(record.id.clone());
                continue;
            }
-            let plugin = load_plugin_definition(
+            let source = describe_install_source(&record.source);
+            match load_plugin_definition(
                &record.install_path,
                record.kind,
-                describe_install_source(&record.source),
+                source.clone(),
                record.kind.marketplace(),
-            )?;
-            if seen_ids.insert(plugin.metadata().id.clone()) {
-                seen_paths.insert(record.install_path.clone());
-                plugins.push(plugin);
+            ) {
+                Ok(plugin) => {
+                    if seen_ids.insert(plugin.metadata().id.clone()) {
+                        seen_paths.insert(record.install_path.clone());
+                        discovery.push_plugin(plugin);
+                    }
+                }
+                Err(error) => {
+                    discovery.push_failure(PluginLoadFailure::new(
+                        record.install_path.clone(),
+                        record.kind,
+                        source,
+                        error,
+                    ));
+                }
            }
        }

@@ -1146,47 +1306,51 @@ impl PluginManager {
            self.store_registry(&registry)?;
        }

-        Ok(plugins)
+        Ok(discovery)
    }

-    fn discover_external_directory_plugins(
+    fn discover_external_directory_plugins_with_failures(
        &self,
        existing_plugins: &[PluginDefinition],
-    ) -> Result<Vec<PluginDefinition>, PluginError> {
-        let mut plugins = Vec::new();
+    ) -> Result<PluginDiscovery, PluginError> {
+        let mut discovery = PluginDiscovery::default();

        for directory in &self.config.external_dirs {
            for root in discover_plugin_dirs(directory)? {
-                let plugin = load_plugin_definition(
+                let source = root.display().to_string();
+                match load_plugin_definition(
                    &root,
                    PluginKind::External,
-                    root.display().to_string(),
+                    source.clone(),
                    EXTERNAL_MARKETPLACE,
-                )?;
-                if existing_plugins
-                    .iter()
-                    .chain(plugins.iter())
-                    .all(|existing| existing.metadata().id != plugin.metadata().id)
-                {
-                    plugins.push(plugin);
+                ) {
+                    Ok(plugin) => {
+                        if existing_plugins
+                            .iter()
+                            .chain(discovery.plugins.iter())
+                            .all(|existing| existing.metadata().id != plugin.metadata().id)
+                        {
+                            discovery.push_plugin(plugin);
+                        }
+                    }
+                    Err(error) => {
+                        discovery.push_failure(PluginLoadFailure::new(
+                            root,
+                            PluginKind::External,
+                            source,
+                            error,
+                        ));
+                    }
                }
            }
        }

-        Ok(plugins)
+        Ok(discovery)
    }

-    fn installed_plugin_registry(&self) -> Result<PluginRegistry, PluginError> {
+    pub fn installed_plugin_registry_report(&self) -> Result<PluginRegistryReport, PluginError> {
        self.sync_bundled_plugins()?;
-        Ok(PluginRegistry::new(
-            self.discover_installed_plugins()?
-                .into_iter()
-                .map(|plugin| {
-                    let enabled = self.is_enabled(plugin.metadata());
-                    RegisteredPlugin::new(plugin, enabled)
-                })
-                .collect(),
-        ))
+        Ok(self.build_registry_report(self.discover_installed_plugins_with_failures()?))
    }

    fn sync_bundled_plugins(&self) -> Result<(), PluginError> {
@@ -1332,6 +1496,26 @@ impl PluginManager {
            }
        })
    }
+
+    fn installed_plugin_registry(&self) -> Result<PluginRegistry, PluginError> {
+        self.installed_plugin_registry_report()?.into_registry()
+    }
+
+    fn build_registry_report(&self, discovery: PluginDiscovery) -> PluginRegistryReport {
+        PluginRegistryReport::new(
+            PluginRegistry::new(
+                discovery
+                    .plugins
+                    .into_iter()
+                    .map(|plugin| {
+                        let enabled = self.is_enabled(plugin.metadata());
+                        RegisteredPlugin::new(plugin, enabled)
+                    })
+                    .collect(),
+            ),
+            discovery.failures,
+        )
+    }
 }

 #[must_use]
@@ -1414,10 +1598,73 @@ fn load_manifest_from_path(
            manifest_path.display()
        ))
    })?;
-    let raw_manifest: RawPluginManifest = serde_json::from_str(&contents)?;
+    let raw_json: Value = serde_json::from_str(&contents)?;
+    let compatibility_errors = detect_claude_code_manifest_contract_gaps(&raw_json);
+    if !compatibility_errors.is_empty() {
+        return Err(PluginError::ManifestValidation(compatibility_errors));
+    }
+    let raw_manifest: RawPluginManifest = serde_json::from_value(raw_json)?;
    build_plugin_manifest(root, raw_manifest)
 }

+fn detect_claude_code_manifest_contract_gaps(
+    raw_manifest: &Value,
+) -> Vec<PluginManifestValidationError> {
+    let Some(root) = raw_manifest.as_object() else {
+        return Vec::new();
+    };
+
+    let mut errors = Vec::new();
+
+    for (field, detail) in [
+        (
+            "skills",
+            "plugin manifest field `skills` uses the Claude Code plugin contract; `claw` does not load plugin-managed skills and instead discovers skills from local roots such as `.claw/skills`, `.omc/skills`, `.agents/skills`, `~/.omc/skills`, and `~/.claude/skills/omc-learned`.",
+        ),
+        (
+            "mcpServers",
+            "plugin manifest field `mcpServers` uses the Claude Code plugin contract; `claw` does not import MCP servers from plugin manifests.",
+        ),
+        (
+            "agents",
+            "plugin manifest field `agents` uses the Claude Code plugin contract; `claw` does not load plugin-managed agent markdown catalogs from plugin manifests.",
+        ),
+    ] {
+        if root.contains_key(field) {
+            errors.push(PluginManifestValidationError::UnsupportedManifestContract {
+                detail: detail.to_string(),
+            });
+        }
+    }
+
+    if root
+        .get("commands")
+        .and_then(Value::as_array)
+        .is_some_and(|commands| commands.iter().any(Value::is_string))
+    {
+        errors.push(PluginManifestValidationError::UnsupportedManifestContract {
+            detail: "plugin manifest field `commands` uses Claude Code-style directory globs; `claw` slash dispatch is still built-in and does not load plugin slash command markdown files.".to_string(),
+        });
+    }
+
+    if let Some(hooks) = root.get("hooks").and_then(Value::as_object) {
+        for hook_name in hooks.keys() {
+            if !matches!(
+                hook_name.as_str(),
+                "PreToolUse" | "PostToolUse" | "PostToolUseFailure"
+            ) {
+                errors.push(PluginManifestValidationError::UnsupportedManifestContract {
+                    detail: format!(
+                        "plugin hook `{hook_name}` uses the Claude Code lifecycle contract; `claw` plugins currently support only PreToolUse, PostToolUse, and PostToolUseFailure."
+                    ),
+                });
+            }
+        }
+    }
+
+    errors
+}
+
 fn plugin_manifest_path(root: &Path) -> Result<PathBuf, PluginError> {
    let direct_path = root.join(MANIFEST_FILE_NAME);
    if direct_path.exists() {
@@ -1449,6 +1696,12 @@ fn build_plugin_manifest(
    let permissions = build_manifest_permissions(&raw.permissions, &mut errors);
    validate_command_entries(root, raw.hooks.pre_tool_use.iter(), "hook", &mut errors);
    validate_command_entries(root, raw.hooks.post_tool_use.iter(), "hook", &mut errors);
+    validate_command_entries(
+        root,
+        raw.hooks.post_tool_use_failure.iter(),
+        "hook",
+        &mut errors,
+    );
    validate_command_entries(
        root,
        raw.lifecycle.init.iter(),
@@ -1676,6 +1929,8 @@ fn validate_command_entry(
    };
    if !path.exists() {
        errors.push(PluginManifestValidationError::MissingPath { kind, path });
+    } else if !path.is_file() {
+        errors.push(PluginManifestValidationError::PathIsDirectory { kind, path });
    }
 }

@@ -1691,6 +1946,11 @@ fn resolve_hooks(root: &Path, hooks: &PluginHooks) -> PluginHooks {
            .iter()
            .map(|entry| resolve_hook_entry(root, entry))
            .collect(),
+        post_tool_use_failure: hooks
+            .post_tool_use_failure
+            .iter()
+            .map(|entry| resolve_hook_entry(root, entry))
+            .collect(),
    }
 }

@@ -1739,7 +1999,12 @@ fn validate_hook_paths(root: Option<&Path>, hooks: &PluginHooks) -> Result<(), P
    let Some(root) = root else {
        return Ok(());
    };
-    for entry in hooks.pre_tool_use.iter().chain(hooks.post_tool_use.iter()) {
+    for entry in hooks
+        .pre_tool_use
+        .iter()
+        .chain(hooks.post_tool_use.iter())
+        .chain(hooks.post_tool_use_failure.iter())
+    {
        validate_command_path(root, entry, "hook")?;
    }
    Ok(())
@@ -1783,6 +2048,12 @@ fn validate_command_path(root: &Path, entry: &str, kind: &str) -> Result<(), Plu
            path.display()
        )));
    }
+    if !path.is_file() {
+        return Err(PluginError::InvalidManifest(format!(
+            "{kind} path `{}` must point to a file",
+            path.display()
+        )));
+    }
    Ok(())
 }

@@ -2094,6 +2365,30 @@ mod tests {
        );
    }

+    fn write_directory_path_plugin(root: &Path, name: &str) {
+        fs::create_dir_all(root.join("hooks").join("pre-dir")).expect("hook dir");
+        fs::create_dir_all(root.join("tools").join("tool-dir")).expect("tool dir");
+        fs::create_dir_all(root.join("commands").join("sync-dir")).expect("command dir");
+        fs::create_dir_all(root.join("lifecycle").join("init-dir")).expect("lifecycle dir");
+        write_file(
+            root.join(MANIFEST_FILE_NAME).as_path(),
+            format!(
+                "{{\n  \"name\": \"{name}\",\n  \"version\": \"1.0.0\",\n  \"description\": \"directory path plugin\",\n  \"hooks\": {{\n    \"PreToolUse\": [\"./hooks/pre-dir\"]\n  }},\n  \"lifecycle\": {{\n    \"Init\": [\"./lifecycle/init-dir\"]\n  }},\n  \"tools\": [\n    {{\n      \"name\": \"dir_tool\",\n      \"description\": \"Directory tool\",\n      \"inputSchema\": {{\"type\": \"object\"}},\n      \"command\": \"./tools/tool-dir\"\n    }}\n  ],\n  \"commands\": [\n    {{\n      \"name\": \"sync\",\n      \"description\": \"Directory command\",\n      \"command\": \"./commands/sync-dir\"\n    }}\n  ]\n}}"
+            )
+            .as_str(),
+        );
+    }
+
+    fn write_broken_failure_hook_plugin(root: &Path, name: &str) {
+        write_file(
+            root.join(MANIFEST_RELATIVE_PATH).as_path(),
+            format!(
+                "{{\n  \"name\": \"{name}\",\n  \"version\": \"1.0.0\",\n  \"description\": \"broken plugin\",\n  \"hooks\": {{\n    \"PostToolUseFailure\": [\"./hooks/missing-failure.sh\"]\n  }}\n}}"
+            )
+            .as_str(),
+        );
+    }
+
    fn write_lifecycle_plugin(root: &Path, name: &str, version: &str) -> PathBuf {
        let log_path = root.join("lifecycle.log");
        write_file(
@@ -2122,7 +2417,7 @@ mod tests {
        let script_path = root.join("tools").join("echo-json.sh");
        write_file(
            &script_path,
-            "#!/bin/sh\nINPUT=$(cat)\nprintf '{\"plugin\":\"%s\",\"tool\":\"%s\",\"input\":%s}\\n' \"$CLAW_PLUGIN_ID\" \"$CLAW_TOOL_NAME\" \"$INPUT\"\n",
+            "#!/bin/sh\nINPUT=$(cat)\nprintf '{\"plugin\":\"%s\",\"tool\":\"%s\",\"input\":%s}\\n' \"$CLAWD_PLUGIN_ID\" \"$CLAWD_TOOL_NAME\" \"$INPUT\"\n",
        );
        #[cfg(unix)]
        {
@@ -2289,6 +2584,37 @@ mod tests {
        let _ = fs::remove_dir_all(root);
    }

+    #[test]
+    fn load_plugin_from_directory_rejects_claude_code_manifest_contracts_with_guidance() {
+        let root = temp_dir("manifest-claude-code-contract");
+        write_file(
+            root.join(MANIFEST_FILE_NAME).as_path(),
+            r#"{
+  "name": "oh-my-claudecode",
+  "version": "4.10.2",
+  "description": "Claude Code plugin manifest",
+  "hooks": {
+    "SessionStart": ["scripts/session-start.mjs"]
+  },
+  "agents": ["agents/*.md"],
+  "commands": ["commands/**/*.md"],
+  "skills": "./skills/",
+  "mcpServers": "./.mcp.json"
+}"#,
+        );
+
+        let error = load_plugin_from_directory(&root)
+            .expect_err("Claude Code plugin manifest should fail with guidance");
+        let rendered = error.to_string();
+        assert!(rendered.contains("field `skills` uses the Claude Code plugin contract"));
+        assert!(rendered.contains("field `mcpServers` uses the Claude Code plugin contract"));
+        assert!(rendered.contains("field `agents` uses the Claude Code plugin contract"));
+        assert!(rendered.contains("field `commands` uses Claude Code-style directory globs"));
+        assert!(rendered.contains("hook `SessionStart` uses the Claude Code lifecycle contract"));
+
+        let _ = fs::remove_dir_all(root);
+    }
+
    #[test]
    fn load_plugin_from_directory_rejects_missing_tool_or_command_paths() {
        let root = temp_dir("manifest-paths");
@@ -2315,6 +2641,90 @@ mod tests {
        let _ = fs::remove_dir_all(root);
    }

+    #[test]
+    fn load_plugin_from_directory_rejects_missing_lifecycle_paths() {
+        // given
+        let root = temp_dir("manifest-lifecycle-paths");
+        write_file(
+            root.join(MANIFEST_FILE_NAME).as_path(),
+            r#"{
+  "name": "missing-lifecycle-paths",
+  "version": "1.0.0",
+  "description": "Missing lifecycle path validation",
+  "lifecycle": {
+    "Init": ["./lifecycle/init.sh"],
+    "Shutdown": ["./lifecycle/shutdown.sh"]
+  }
+}"#,
+        );
+
+        // when
+        let error =
+            load_plugin_from_directory(&root).expect_err("missing lifecycle paths should fail");
+
+        // then
+        match error {
+            PluginError::ManifestValidation(errors) => {
+                assert!(errors.iter().any(|error| matches!(
+                    error,
+                    PluginManifestValidationError::MissingPath { kind, path }
+                    if *kind == "lifecycle command"
+                        && path.ends_with(Path::new("lifecycle/init.sh"))
+                )));
+                assert!(errors.iter().any(|error| matches!(
+                    error,
+                    PluginManifestValidationError::MissingPath { kind, path }
+                    if *kind == "lifecycle command"
+                        && path.ends_with(Path::new("lifecycle/shutdown.sh"))
+                )));
+            }
+            other => panic!("expected manifest validation errors, got {other}"),
+        }
+
+        let _ = fs::remove_dir_all(root);
+    }
+
+    #[test]
+    fn load_plugin_from_directory_rejects_directory_command_paths() {
+        // given
+        let root = temp_dir("manifest-directory-paths");
+        write_directory_path_plugin(&root, "directory-paths");
+
+        // when
+        let error =
+            load_plugin_from_directory(&root).expect_err("directory command paths should fail");
+
+        // then
+        match error {
+            PluginError::ManifestValidation(errors) => {
+                assert!(errors.iter().any(|error| matches!(
+                    error,
+                    PluginManifestValidationError::PathIsDirectory { kind, path }
+                    if *kind == "hook" && path.ends_with(Path::new("hooks/pre-dir"))
+                )));
+                assert!(errors.iter().any(|error| matches!(
+                    error,
+                    PluginManifestValidationError::PathIsDirectory { kind, path }
+                    if *kind == "lifecycle command"
+                        && path.ends_with(Path::new("lifecycle/init-dir"))
+                )));
+                assert!(errors.iter().any(|error| matches!(
+                    error,
+                    PluginManifestValidationError::PathIsDirectory { kind, path }
+                    if *kind == "tool" && path.ends_with(Path::new("tools/tool-dir"))
+                )));
+                assert!(errors.iter().any(|error| matches!(
+                    error,
+                    PluginManifestValidationError::PathIsDirectory { kind, path }
+                    if *kind == "command" && path.ends_with(Path::new("commands/sync-dir"))
+                )));
+            }
+            other => panic!("expected manifest validation errors, got {other}"),
+        }
+
+        let _ = fs::remove_dir_all(root);
+    }
+
    #[test]
    fn load_plugin_from_directory_rejects_invalid_permissions() {
        let root = temp_dir("manifest-invalid-permissions");
@@ -2806,16 +3216,95 @@ mod tests {
        let _ = fs::remove_dir_all(source_root);
    }

+    #[test]
+    fn plugin_registry_report_collects_load_failures_without_dropping_valid_plugins() {
+        // given
+        let config_home = temp_dir("report-home");
+        let external_root = temp_dir("report-external");
+        write_external_plugin(&external_root.join("valid"), "valid-report", "1.0.0");
+        write_broken_plugin(&external_root.join("broken"), "broken-report");
+
+        let mut config = PluginManagerConfig::new(&config_home);
+        config.external_dirs = vec![external_root.clone()];
+        let manager = PluginManager::new(config);
+
+        // when
+        let report = manager
+            .plugin_registry_report()
+            .expect("report should tolerate invalid external plugins");
+
+        // then
+        assert!(report.registry().contains("valid-report@external"));
+        assert_eq!(report.failures().len(), 1);
+        assert_eq!(report.failures()[0].kind, PluginKind::External);
+        assert!(report.failures()[0]
+            .plugin_root
+            .ends_with(Path::new("broken")));
+        assert!(report.failures()[0]
+            .error()
+            .to_string()
+            .contains("does not exist"));
+
+        let error = manager
+            .plugin_registry()
+            .expect_err("strict registry should surface load failures");
+        match error {
+            PluginError::LoadFailures(failures) => {
+                assert_eq!(failures.len(), 1);
+                assert!(failures[0].plugin_root.ends_with(Path::new("broken")));
+            }
+            other => panic!("expected load failures, got {other}"),
+        }
+
+        let _ = fs::remove_dir_all(config_home);
+        let _ = fs::remove_dir_all(external_root);
+    }
+
+    #[test]
+    fn installed_plugin_registry_report_collects_load_failures_from_install_root() {
+        // given
+        let config_home = temp_dir("installed-report-home");
+        let bundled_root = temp_dir("installed-report-bundled");
+        let install_root = config_home.join("plugins").join("installed");
+        write_external_plugin(&install_root.join("valid"), "installed-valid", "1.0.0");
+        write_broken_plugin(&install_root.join("broken"), "installed-broken");
+
+        let mut config = PluginManagerConfig::new(&config_home);
+        config.bundled_root = Some(bundled_root.clone());
+        config.install_root = Some(install_root);
+        let manager = PluginManager::new(config);
+
+        // when
+        let report = manager
+            .installed_plugin_registry_report()
+            .expect("installed report should tolerate invalid installed plugins");
+
+        // then
+        assert!(report.registry().contains("installed-valid@external"));
+        assert_eq!(report.failures().len(), 1);
+        assert!(report.failures()[0]
+            .plugin_root
+            .ends_with(Path::new("broken")));
+
+        let _ = fs::remove_dir_all(config_home);
+        let _ = fs::remove_dir_all(bundled_root);
+    }
+
    #[test]
    fn rejects_plugin_sources_with_missing_hook_paths() {
+        // given
        let config_home = temp_dir("broken-home");
        let source_root = temp_dir("broken-source");
        write_broken_plugin(&source_root, "broken");

        let manager = PluginManager::new(PluginManagerConfig::new(&config_home));
+
+        // when
        let error = manager
            .validate_plugin_source(source_root.to_str().expect("utf8 path"))
            .expect_err("missing hook file should fail validation");
+
+        // then
        assert!(error.to_string().contains("does not exist"));

        let mut manager = PluginManager::new(PluginManagerConfig::new(&config_home));
@@ -2828,6 +3317,33 @@ mod tests {
        let _ = fs::remove_dir_all(source_root);
    }

+    #[test]
+    fn rejects_plugin_sources_with_missing_failure_hook_paths() {
+        // given
+        let config_home = temp_dir("broken-failure-home");
+        let source_root = temp_dir("broken-failure-source");
+        write_broken_failure_hook_plugin(&source_root, "broken-failure");
+
+        let manager = PluginManager::new(PluginManagerConfig::new(&config_home));
+
+        // when
+        let error = manager
+            .validate_plugin_source(source_root.to_str().expect("utf8 path"))
+            .expect_err("missing failure hook file should fail validation");
+
+        // then
+        assert!(error.to_string().contains("does not exist"));
+
+        let mut manager = PluginManager::new(PluginManagerConfig::new(&config_home));
+        let install_error = manager
+            .install(source_root.to_str().expect("utf8 path"))
+            .expect_err("install should reject invalid failure hook paths");
+        assert!(install_error.to_string().contains("does not exist"));
+
+        let _ = fs::remove_dir_all(config_home);
+        let _ = fs::remove_dir_all(source_root);
+    }
+
    #[test]
    fn plugin_registry_runs_initialize_and_shutdown_for_enabled_plugins() {
        let config_home = temp_dir("lifecycle-home");
@@ -7,13 +7,14 @@ publish.workspace = true

 [dependencies]
 sha2 = "0.10"
+which = "7"
 glob = "0.3"
-lsp = { path = "../lsp" }
 plugins = { path = "../plugins" }
 regex = "1"
 serde = { version = "1", features = ["derive"] }
 serde_json.workspace = true
-tokio = { version = "1", features = ["io-util", "macros", "process", "rt", "rt-multi-thread", "time"] }
+telemetry = { path = "../telemetry" }
+tokio = { version = "1", features = ["io-std", "io-util", "macros", "process", "rt", "rt-multi-thread", "time"] }
 walkdir = "2"

 [lints]
@@ -0,0 +1,60 @@
+# 运行时模块 (runtime)
+
+本模块是 Claw 的核心引擎，负责协调 AI 模型、工具执行、会话管理和权限控制之间的所有交互。
+
+## 概览
+
+`runtime` 模块是整个系统的“中枢神经”，其主要职责包括：
+- **对话驱动**：管理“用户-助手-工具”的循环迭代。
+- **会话持久化**：负责会话的加载、保存及历史记录的压缩 (Compaction)。
+- **MCP 客户端**：实现模型上下文协议 (Model Context Protocol)，支持与外部 MCP 服务器通信。
+- **安全沙箱与权限**：实施基于策略的工具执行权限检查。
+- **上下文构建**：动态生成系统提示词 (System Prompt)，集成工作区上下文。
+- **消耗统计**：精确跟踪 Token 使用情况和 Token 缓存状态。
+
+## 关键特性
+
+- **ConversationRuntime**：核心驱动类，支持流式响应处理和多轮工具调用迭代。
+- **权限引擎 (Permissions)**：提供多种模式（`ReadOnly`, `WorkspaceWrite`, `DangerFullAccess`），并支持交互式权限确认。
+- **会话压缩 (Compaction)**：当对话历史过长影响性能或成本时，自动将旧消息总结为摘要，保持上下文精简。
+- **钩子集成 (Hooks)**：在工具执行的前后触发插件定义的钩子，支持干预工具输入或处理执行结果。
+- **沙箱执行 (Sandbox)**：为 Bash 等敏感工具提供受限的执行环境。
+
+## 实现逻辑
+
+### 核心子模块
+
+- **`conversation.rs`**: 定义了核心的 `ConversationRuntime` 和 `ApiClient`/`ToolExecutor` 特性。
+- **`mcp_stdio.rs` / `mcp_client.rs`**: 实现了完整的 MCP 规范，支持通过标准输入/输出与外部工具服务器交互。
+- **`session.rs`**: 定义了消息模型 (`ConversationMessage`)、内容块 (`ContentBlock`) 和会话序列化逻辑。
+- **`permissions.rs`**: 实现了权限审核逻辑和提示器接口。
+- **`compact.rs`**: 包含了基于 LLM 的会话摘要生成和历史裁剪算法。
+- **`config.rs`**: 负责加载和合并多层级的配置文件。
+
+### 对话循环流程 (run_turn)
+
+1. 将用户输入推入 `Session`。
+2. 调用 `ApiClient` 发起流式请求。
+3. 监听 `AssistantEvent`，解析文本内容和工具调用请求。
+4. **工具权限审核**：针对每个 `ToolUse`，根据 `PermissionPolicy` 决定是允许、拒绝还是询问用户。
+5. **执行工具**：若允许，则通过 `ToolExecutor`（或 MCP 客户端）执行工具，并运行相关的 `Pre/Post Hooks`。
+6. 将工具结果反馈给 AI，进入下一轮迭代，直到 AI 给出最终回复。
+
+## 使用示例 (内部)
+
+```rust
+use runtime::{ConversationRuntime, Session, PermissionPolicy, PermissionMode};
+
+// 初始化运行时
+let mut runtime = ConversationRuntime::new(
+    Session::new(),
+    api_client,
+    tool_executor,
+    PermissionPolicy::new(PermissionMode::WorkspaceWrite),
+    system_prompt,
+);
+
+// 运行一轮对话
+let summary = runtime.run_turn("帮我重构 src/lib.rs", Some(&mut cli_prompter))?;
+println!("共迭代 {} 次，消耗 {} tokens", summary.iterations, summary.usage.total_tokens());
+```
@@ -14,6 +14,7 @@ use crate::sandbox::{
 };
 use crate::ConfigLoader;

+/// Input schema for the built-in bash execution tool.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct BashCommandInput {
    pub command: String,
@@ -33,6 +34,7 @@ pub struct BashCommandInput {
    pub allowed_mounts: Option<Vec<String>>,
 }

+/// Output returned from a bash tool invocation.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 pub struct BashCommandOutput {
    pub stdout: String,
@@ -64,6 +66,7 @@ pub struct BashCommandOutput {
    pub sandbox_status: Option<SandboxStatus>,
 }

+/// Executes a shell command with the requested sandbox settings.
 pub fn execute_bash(input: BashCommandInput) -> io::Result<BashCommandOutput> {
    let cwd = env::current_dir()?;
    let sandbox_status = sandbox_status_for_input(&input, &cwd);
@@ -134,8 +137,8 @@ async fn execute_bash_async(
    };

    let (output, interrupted) = output_result;
-    let stdout = String::from_utf8_lossy(&output.stdout).into_owned();
-    let stderr = String::from_utf8_lossy(&output.stderr).into_owned();
+    let stdout = truncate_output(&String::from_utf8_lossy(&output.stdout));
+    let stderr = truncate_output(&String::from_utf8_lossy(&output.stderr));
    let no_output_expected = Some(stdout.trim().is_empty() && stderr.trim().is_empty());
    let return_code_interpretation = output.status.code().and_then(|code| {
        if code == 0 {
@@ -197,36 +200,31 @@ fn prepare_command(
        return prepared;
    }

-    let mut prepared = if cfg!(target_os = "windows") && !sh_exists() {
+    let prepared = if cfg!(target_os = "windows") && !sh_exists() {
        let mut p = Command::new("cmd");
-        p.arg("/C").arg(command);
+        p.arg("/C").arg(command).current_dir(cwd);
        p
    } else {
        let mut p = Command::new("sh");
-        p.arg("-lc").arg(command);
+        p.arg("-lc").arg(command).current_dir(cwd);
+        if sandbox_status.filesystem_active {
+            p.env("HOME", cwd.join(".sandbox-home"));
+            p.env("TMPDIR", cwd.join(".sandbox-tmp"));
+        }
        p
    };
-    prepared.current_dir(cwd);
-    if sandbox_status.filesystem_active {
-        prepared.env("HOME", cwd.join(".sandbox-home"));
-        prepared.env("TMPDIR", cwd.join(".sandbox-tmp"));
-    }
    prepared
 }

 fn sh_exists() -> bool {
-    env::var_os("PATH").is_some_and(|paths| {
-        env::split_paths(&paths).any(|path| {
-            #[cfg(windows)]
-            {
-                path.join("sh.exe").exists() || path.join("sh.bat").exists() || path.join("sh").exists()
-            }
-            #[cfg(not(windows))]
-            {
-                path.join("sh").exists()
-            }
-        })
-    })
+    #[cfg(windows)]
+    {
+        which::which("sh").is_ok()
+    }
+    #[cfg(not(windows))]
+    {
+        true
+    }
 }

 fn prepare_tokio_command(
@@ -247,20 +245,19 @@ fn prepare_tokio_command(
        return prepared;
    }

-    let mut prepared = if cfg!(target_os = "windows") && !sh_exists() {
+    let prepared = if cfg!(target_os = "windows") && !sh_exists() {
        let mut p = TokioCommand::new("cmd");
-        p.arg("/C").arg(command);
+        p.arg("/C").arg(command).current_dir(cwd);
        p
    } else {
        let mut p = TokioCommand::new("sh");
-        p.arg("-lc").arg(command);
+        p.arg("-lc").arg(command).current_dir(cwd);
+        if sandbox_status.filesystem_active {
+            p.env("HOME", cwd.join(".sandbox-home"));
+            p.env("TMPDIR", cwd.join(".sandbox-tmp"));
+        }
        p
    };
-    prepared.current_dir(cwd);
-    if sandbox_status.filesystem_active {
-        prepared.env("HOME", cwd.join(".sandbox-home"));
-        prepared.env("TMPDIR", cwd.join(".sandbox-tmp"));
-    }
    prepared
 }

@@ -312,3 +309,53 @@ mod tests {
        assert!(!output.sandbox_status.expect("sandbox status").enabled);
    }
 }
+
+/// Maximum output bytes before truncation (16 KiB, matching upstream).
+const MAX_OUTPUT_BYTES: usize = 16_384;
+
+/// Truncate output to `MAX_OUTPUT_BYTES`, appending a marker when trimmed.
+fn truncate_output(s: &str) -> String {
+    if s.len() <= MAX_OUTPUT_BYTES {
+        return s.to_string();
+    }
+    // Find the last valid UTF-8 boundary at or before MAX_OUTPUT_BYTES
+    let mut end = MAX_OUTPUT_BYTES;
+    while end > 0 && !s.is_char_boundary(end) {
+        end -= 1;
+    }
+    let mut truncated = s[..end].to_string();
+    truncated.push_str("\n\n[output truncated — exceeded 16384 bytes]");
+    truncated
+}
+
+#[cfg(test)]
+mod truncation_tests {
+    use super::*;
+
+    #[test]
+    fn short_output_unchanged() {
+        let s = "hello world";
+        assert_eq!(truncate_output(s), s);
+    }
+
+    #[test]
+    fn long_output_truncated() {
+        let s = "x".repeat(20_000);
+        let result = truncate_output(&s);
+        assert!(result.len() < 20_000);
+        assert!(result.ends_with("[output truncated — exceeded 16384 bytes]"));
+    }
+
+    #[test]
+    fn exact_boundary_unchanged() {
+        let s = "a".repeat(MAX_OUTPUT_BYTES);
+        assert_eq!(truncate_output(&s), s);
+    }
+
+    #[test]
+    fn one_over_boundary_truncated() {
+        let s = "a".repeat(MAX_OUTPUT_BYTES + 1);
+        let result = truncate_output(&s);
+        assert!(result.contains("[output truncated"));
+    }
+}
@@ -21,7 +21,7 @@ pub struct BootstrapPlan {

 impl BootstrapPlan {
    #[must_use]
-    pub fn claw_default() -> Self {
+    pub fn claude_code_default() -> Self {
        Self::from_phases(vec![
            BootstrapPhase::CliEntry,
            BootstrapPhase::FastPathVersion,
@@ -54,3 +54,58 @@ impl BootstrapPlan {
        &self.phases
    }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::{BootstrapPhase, BootstrapPlan};
+
+    #[test]
+    fn from_phases_deduplicates_while_preserving_order() {
+        // given
+        let phases = vec![
+            BootstrapPhase::CliEntry,
+            BootstrapPhase::FastPathVersion,
+            BootstrapPhase::CliEntry,
+            BootstrapPhase::MainRuntime,
+            BootstrapPhase::FastPathVersion,
+        ];
+
+        // when
+        let plan = BootstrapPlan::from_phases(phases);
+
+        // then
+        assert_eq!(
+            plan.phases(),
+            &[
+                BootstrapPhase::CliEntry,
+                BootstrapPhase::FastPathVersion,
+                BootstrapPhase::MainRuntime,
+            ]
+        );
+    }
+
+    #[test]
+    fn claude_code_default_covers_each_phase_once() {
+        // given
+        let expected = [
+            BootstrapPhase::CliEntry,
+            BootstrapPhase::FastPathVersion,
+            BootstrapPhase::StartupProfiler,
+            BootstrapPhase::SystemPromptFastPath,
+            BootstrapPhase::ChromeMcpFastPath,
+            BootstrapPhase::DaemonWorkerFastPath,
+            BootstrapPhase::BridgeFastPath,
+            BootstrapPhase::DaemonFastPath,
+            BootstrapPhase::BackgroundSessionFastPath,
+            BootstrapPhase::TemplateFastPath,
+            BootstrapPhase::EnvironmentRunnerFastPath,
+            BootstrapPhase::MainRuntime,
+        ];
+
+        // when
+        let plan = BootstrapPlan::claude_code_default();
+
+        // then
+        assert_eq!(plan.phases(), &expected);
+    }
+}
@@ -0,0 +1,144 @@
+use serde::{Deserialize, Serialize};
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct BranchLockIntent {
+    #[serde(rename = "laneId")]
+    pub lane_id: String,
+    pub branch: String,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub worktree: Option<String>,
+    #[serde(default, skip_serializing_if = "Vec::is_empty")]
+    pub modules: Vec<String>,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct BranchLockCollision {
+    pub branch: String,
+    pub module: String,
+    #[serde(rename = "laneIds")]
+    pub lane_ids: Vec<String>,
+}
+
+#[must_use]
+pub fn detect_branch_lock_collisions(intents: &[BranchLockIntent]) -> Vec<BranchLockCollision> {
+    let mut collisions = Vec::new();
+
+    for (index, left) in intents.iter().enumerate() {
+        for right in &intents[index + 1..] {
+            if left.branch != right.branch {
+                continue;
+            }
+            for module in overlapping_modules(&left.modules, &right.modules) {
+                collisions.push(BranchLockCollision {
+                    branch: left.branch.clone(),
+                    module,
+                    lane_ids: vec![left.lane_id.clone(), right.lane_id.clone()],
+                });
+            }
+        }
+    }
+
+    collisions.sort_by(|a, b| {
+        a.branch
+            .cmp(&b.branch)
+            .then(a.module.cmp(&b.module))
+            .then(a.lane_ids.cmp(&b.lane_ids))
+    });
+    collisions.dedup();
+    collisions
+}
+
+fn overlapping_modules(left: &[String], right: &[String]) -> Vec<String> {
+    let mut overlaps = Vec::new();
+    for left_module in left {
+        for right_module in right {
+            if modules_overlap(left_module, right_module) {
+                overlaps.push(shared_scope(left_module, right_module));
+            }
+        }
+    }
+    overlaps.sort();
+    overlaps.dedup();
+    overlaps
+}
+
+fn modules_overlap(left: &str, right: &str) -> bool {
+    left == right
+        || left.starts_with(&format!("{right}/"))
+        || right.starts_with(&format!("{left}/"))
+}
+
+fn shared_scope(left: &str, right: &str) -> String {
+    if left.starts_with(&format!("{right}/")) || left == right {
+        right.to_string()
+    } else {
+        left.to_string()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{detect_branch_lock_collisions, BranchLockIntent};
+
+    #[test]
+    fn detects_same_branch_same_module_collisions() {
+        let collisions = detect_branch_lock_collisions(&[
+            BranchLockIntent {
+                lane_id: "lane-a".to_string(),
+                branch: "feature/lock".to_string(),
+                worktree: Some("wt-a".to_string()),
+                modules: vec!["runtime/mcp".to_string()],
+            },
+            BranchLockIntent {
+                lane_id: "lane-b".to_string(),
+                branch: "feature/lock".to_string(),
+                worktree: Some("wt-b".to_string()),
+                modules: vec!["runtime/mcp".to_string()],
+            },
+        ]);
+
+        assert_eq!(collisions.len(), 1);
+        assert_eq!(collisions[0].branch, "feature/lock");
+        assert_eq!(collisions[0].module, "runtime/mcp");
+    }
+
+    #[test]
+    fn detects_nested_module_scope_collisions() {
+        let collisions = detect_branch_lock_collisions(&[
+            BranchLockIntent {
+                lane_id: "lane-a".to_string(),
+                branch: "feature/lock".to_string(),
+                worktree: None,
+                modules: vec!["runtime".to_string()],
+            },
+            BranchLockIntent {
+                lane_id: "lane-b".to_string(),
+                branch: "feature/lock".to_string(),
+                worktree: None,
+                modules: vec!["runtime/mcp".to_string()],
+            },
+        ]);
+
+        assert_eq!(collisions[0].module, "runtime");
+    }
+
+    #[test]
+    fn ignores_different_branches() {
+        let collisions = detect_branch_lock_collisions(&[
+            BranchLockIntent {
+                lane_id: "lane-a".to_string(),
+                branch: "feature/a".to_string(),
+                worktree: None,
+                modules: vec!["runtime/mcp".to_string()],
+            },
+            BranchLockIntent {
+                lane_id: "lane-b".to_string(),
+                branch: "feature/b".to_string(),
+                worktree: None,
+                modules: vec!["runtime/mcp".to_string()],
+            },
+        ]);
+
+        assert!(collisions.is_empty());
+    }
+}
@@ -5,6 +5,7 @@ const COMPACT_CONTINUATION_PREAMBLE: &str =
 const COMPACT_RECENT_MESSAGES_NOTE: &str = "Recent messages are preserved verbatim.";
 const COMPACT_DIRECT_RESUME_INSTRUCTION: &str = "Continue the conversation from where it left off without asking the user any further questions. Resume directly — do not acknowledge the summary, do not recap what was happening, and do not preface with continuation text.";

+/// Thresholds controlling when and how a session is compacted.
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub struct CompactionConfig {
    pub preserve_recent_messages: usize,
@@ -20,6 +21,7 @@ impl Default for CompactionConfig {
    }
 }

+/// Result of compacting a session into a summary plus preserved tail messages.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct CompactionResult {
    pub summary: String,
@@ -28,11 +30,13 @@ pub struct CompactionResult {
    pub removed_message_count: usize,
 }

+/// Roughly estimates the token footprint of the current session transcript.
 #[must_use]
 pub fn estimate_session_tokens(session: &Session) -> usize {
    session.messages.iter().map(estimate_message_tokens).sum()
 }

+/// Returns `true` when the session exceeds the configured compaction budget.
 #[must_use]
 pub fn should_compact(session: &Session, config: CompactionConfig) -> bool {
    let start = compacted_summary_prefix_len(session);
@@ -46,6 +50,7 @@ pub fn should_compact(session: &Session, config: CompactionConfig) -> bool {
            >= config.max_estimated_tokens
 }

+/// Normalizes a compaction summary into user-facing continuation text.
 #[must_use]
 pub fn format_compact_summary(summary: &str) -> String {
    let without_analysis = strip_tag_block(summary, "analysis");
@@ -61,6 +66,7 @@ pub fn format_compact_summary(summary: &str) -> String {
    collapse_blank_lines(&formatted).trim().to_string()
 }

+/// Builds the synthetic system message used after session compaction.
 #[must_use]
 pub fn get_compact_continuation_message(
    summary: &str,
@@ -85,6 +91,7 @@ pub fn get_compact_continuation_message(
    base
 }

+/// Compacts a session by summarizing older messages and preserving the recent tail.
 #[must_use]
 pub fn compact_session(session: &Session, config: CompactionConfig) -> CompactionResult {
    if !should_compact(session, config) {
@@ -101,10 +108,55 @@ pub fn compact_session(session: &Session, config: CompactionConfig) -> Compactio
        .first()
        .and_then(extract_existing_compacted_summary);
    let compacted_prefix_len = usize::from(existing_summary.is_some());
-    let keep_from = session
+    let raw_keep_from = session
        .messages
        .len()
        .saturating_sub(config.preserve_recent_messages);
+    // Ensure we do not split a tool-use / tool-result pair at the compaction
+    // boundary. If the first preserved message is a user message whose first
+    // block is a ToolResult, the assistant message with the matching ToolUse
+    // was slated for removal — that produces an orphaned tool role message on
+    // the OpenAI-compat path (400: tool message must follow assistant with
+    // tool_calls). Walk the boundary back until we start at a safe point.
+    let keep_from = {
+        let mut k = raw_keep_from;
+        // If the first preserved message is a tool-result turn, ensure its
+        // paired assistant tool-use turn is preserved too. Without this fix,
+        // the OpenAI-compat adapter sends an orphaned 'tool' role message
+        // with no preceding assistant 'tool_calls', which providers reject
+        // with a 400. We walk back only if the immediately preceding message
+        // is NOT an assistant message that contains a ToolUse block (i.e. the
+        // pair is actually broken at the boundary).
+        loop {
+            if k == 0 || k <= compacted_prefix_len {
+                break;
+            }
+            let first_preserved = &session.messages[k];
+            let starts_with_tool_result = first_preserved
+                .blocks
+                .first()
+                .map(|b| matches!(b, ContentBlock::ToolResult { .. }))
+                .unwrap_or(false);
+            if !starts_with_tool_result {
+                break;
+            }
+            // Check the message just before the current boundary.
+            let preceding = &session.messages[k - 1];
+            let preceding_has_tool_use = preceding
+                .blocks
+                .iter()
+                .any(|b| matches!(b, ContentBlock::ToolUse { .. }));
+            if preceding_has_tool_use {
+                // Pair is intact — walk back one more to include the assistant turn.
+                k = k.saturating_sub(1);
+                break;
+            }
+            // Preceding message has no ToolUse but we have a ToolResult —
+            // this is already an orphaned pair; walk back to try to fix it.
+            k = k.saturating_sub(1);
+        }
+        k
+    };
    let removed = &session.messages[compacted_prefix_len..keep_from];
    let preserved = session.messages[keep_from..].to_vec();
    let summary =
@@ -119,13 +171,14 @@ pub fn compact_session(session: &Session, config: CompactionConfig) -> Compactio
    }];
    compacted_messages.extend(preserved);

+    let mut compacted_session = session.clone();
+    compacted_session.messages = compacted_messages;
+    compacted_session.record_compaction(summary.clone(), removed.len());
+
    CompactionResult {
        summary,
        formatted_summary,
-        compacted_session: Session {
-            version: session.version,
-            messages: compacted_messages,
-        },
+        compacted_session,
        removed_message_count: removed.len(),
    }
 }
@@ -502,7 +555,7 @@ fn extract_summary_timeline(summary: &str) -> Vec<String> {
 #[cfg(test)]
 mod tests {
    use super::{
-        collect_key_files, compact_session, estimate_session_tokens, format_compact_summary,
+        collect_key_files, compact_session, format_compact_summary,
        get_compact_continuation_message, infer_pending_work, should_compact, CompactionConfig,
    };
    use crate::session::{ContentBlock, ConversationMessage, MessageRole, Session};
@@ -515,10 +568,8 @@ mod tests {

    #[test]
    fn leaves_small_sessions_unchanged() {
-        let session = Session {
-            version: 1,
-            messages: vec![ConversationMessage::user_text("hello")],
-        };
+        let mut session = Session::new();
+        session.messages = vec![ConversationMessage::user_text("hello")];

        let result = compact_session(&session, CompactionConfig::default());
        assert_eq!(result.removed_message_count, 0);
@@ -529,23 +580,21 @@ mod tests {

    #[test]
    fn compacts_older_messages_into_a_system_summary() {
-        let session = Session {
-            version: 1,
-            messages: vec![
-                ConversationMessage::user_text("one ".repeat(200)),
-                ConversationMessage::assistant(vec![ContentBlock::Text {
-                    text: "two ".repeat(200),
-                }]),
-                ConversationMessage::tool_result("1", "bash", "ok ".repeat(200), false),
-                ConversationMessage {
-                    role: MessageRole::Assistant,
-                    blocks: vec![ContentBlock::Text {
-                        text: "recent".to_string(),
-                    }],
-                    usage: None,
-                },
-            ],
-        };
+        let mut session = Session::new();
+        session.messages = vec![
+            ConversationMessage::user_text("one ".repeat(200)),
+            ConversationMessage::assistant(vec![ContentBlock::Text {
+                text: "two ".repeat(200),
+            }]),
+            ConversationMessage::tool_result("1", "bash", "ok ".repeat(200), false),
+            ConversationMessage {
+                role: MessageRole::Assistant,
+                blocks: vec![ContentBlock::Text {
+                    text: "recent".to_string(),
+                }],
+                usage: None,
+            },
+        ];

        let result = compact_session(
            &session,
@@ -555,7 +604,14 @@ mod tests {
            },
        );

-        assert_eq!(result.removed_message_count, 2);
+        // With the tool-use/tool-result boundary fix, the compaction preserves
+        // one extra message to avoid an orphaned tool result at the boundary.
+        // messages[1] (assistant) must be kept along with messages[2] (tool result).
+        assert!(
+            result.removed_message_count <= 2,
+            "expected at most 2 removed, got {}",
+            result.removed_message_count
+        );
        assert_eq!(
            result.compacted_session.messages[0].role,
            MessageRole::System
@@ -573,28 +629,29 @@ mod tests {
                max_estimated_tokens: 1,
            }
        ));
+        // Note: with the tool-use/tool-result boundary guard the compacted session
+        // may preserve one extra message at the boundary, so token reduction is
+        // not guaranteed for small sessions. The invariant that matters is that
+        // the removed_message_count is non-zero (something was compacted).
        assert!(
-            estimate_session_tokens(&result.compacted_session) < estimate_session_tokens(&session)
+            result.removed_message_count > 0,
+            "compaction must remove at least one message"
        );
    }

    #[test]
    fn keeps_previous_compacted_context_when_compacting_again() {
-        let initial_session = Session {
-            version: 1,
-            messages: vec![
-                ConversationMessage::user_text("Investigate rust/crates/runtime/src/compact.rs"),
-                ConversationMessage::assistant(vec![ContentBlock::Text {
-                    text: "I will inspect the compact flow.".to_string(),
-                }]),
-                ConversationMessage::user_text(
-                    "Also update rust/crates/runtime/src/conversation.rs",
-                ),
-                ConversationMessage::assistant(vec![ContentBlock::Text {
-                    text: "Next: preserve prior summary context during auto compact.".to_string(),
-                }]),
-            ],
-        };
+        let mut initial_session = Session::new();
+        initial_session.messages = vec![
+            ConversationMessage::user_text("Investigate rust/crates/runtime/src/compact.rs"),
+            ConversationMessage::assistant(vec![ContentBlock::Text {
+                text: "I will inspect the compact flow.".to_string(),
+            }]),
+            ConversationMessage::user_text("Also update rust/crates/runtime/src/conversation.rs"),
+            ConversationMessage::assistant(vec![ContentBlock::Text {
+                text: "Next: preserve prior summary context during auto compact.".to_string(),
+            }]),
+        ];
        let config = CompactionConfig {
            preserve_recent_messages: 2,
            max_estimated_tokens: 1,
@@ -609,13 +666,9 @@ mod tests {
            }]),
        ]);

-        let second = compact_session(
-            &Session {
-                version: 1,
-                messages: follow_up_messages,
-            },
-            config,
-        );
+        let mut second_session = Session::new();
+        second_session.messages = follow_up_messages;
+        let second = compact_session(&second_session, config);

        assert!(second
            .formatted_summary
@@ -644,22 +697,20 @@ mod tests {
    #[test]
    fn ignores_existing_compacted_summary_when_deciding_to_recompact() {
        let summary = "<summary>Conversation summary:\n- Scope: earlier work preserved.\n- Key timeline:\n  - user: large preserved context\n</summary>";
-        let session = Session {
-            version: 1,
-            messages: vec![
-                ConversationMessage {
-                    role: MessageRole::System,
-                    blocks: vec![ContentBlock::Text {
-                        text: get_compact_continuation_message(summary, true, true),
-                    }],
-                    usage: None,
-                },
-                ConversationMessage::user_text("tiny"),
-                ConversationMessage::assistant(vec![ContentBlock::Text {
-                    text: "recent".to_string(),
-                }]),
-            ],
-        };
+        let mut session = Session::new();
+        session.messages = vec![
+            ConversationMessage {
+                role: MessageRole::System,
+                blocks: vec![ContentBlock::Text {
+                    text: get_compact_continuation_message(summary, true, true),
+                }],
+                usage: None,
+            },
+            ConversationMessage::user_text("tiny"),
+            ConversationMessage::assistant(vec![ContentBlock::Text {
+                text: "recent".to_string(),
+            }]),
+        ];

        assert!(!should_compact(
            &session,
@@ -682,10 +733,84 @@ mod tests {
    #[test]
    fn extracts_key_files_from_message_content() {
        let files = collect_key_files(&[ConversationMessage::user_text(
-            "Update rust/crates/runtime/src/compact.rs and rust/crates/tools/src/lib.rs next.",
+            "Update rust/crates/runtime/src/compact.rs and rust/crates/rusty-claude-cli/src/main.rs next.",
        )]);
        assert!(files.contains(&"rust/crates/runtime/src/compact.rs".to_string()));
-        assert!(files.contains(&"rust/crates/tools/src/lib.rs".to_string()));
+        assert!(files.contains(&"rust/crates/rusty-claude-cli/src/main.rs".to_string()));
+    }
+
+    /// Regression: compaction must not split an assistant(ToolUse) /
+    /// user(ToolResult) pair at the boundary. An orphaned tool-result message
+    /// without the preceding assistant tool_calls causes a 400 on the
+    /// OpenAI-compat path (gaebal-gajae repro 2026-04-09).
+    #[test]
+    fn compaction_does_not_split_tool_use_tool_result_pair() {
+        use crate::session::{ContentBlock, Session};
+
+        let tool_id = "call_abc";
+        let mut session = Session::default();
+        // Turn 1: user prompt
+        session
+            .push_message(ConversationMessage::user_text("Search for files"))
+            .unwrap();
+        // Turn 2: assistant calls a tool
+        session
+            .push_message(ConversationMessage::assistant(vec![
+                ContentBlock::ToolUse {
+                    id: tool_id.to_string(),
+                    name: "search".to_string(),
+                    input: "{\"q\":\"*.rs\"}".to_string(),
+                },
+            ]))
+            .unwrap();
+        // Turn 3: tool result
+        session
+            .push_message(ConversationMessage::tool_result(
+                tool_id,
+                "search",
+                "found 5 files",
+                false,
+            ))
+            .unwrap();
+        // Turn 4: assistant final response
+        session
+            .push_message(ConversationMessage::assistant(vec![ContentBlock::Text {
+                text: "Done.".to_string(),
+            }]))
+            .unwrap();
+
+        // Compact preserving only 1 recent message — without the fix this
+        // would cut the boundary so that the tool result (turn 3) is first,
+        // without its preceding assistant tool_calls (turn 2).
+        let config = CompactionConfig {
+            preserve_recent_messages: 1,
+            ..CompactionConfig::default()
+        };
+        let result = compact_session(&session, config);
+        // After compaction, no two consecutive messages should have the pattern
+        // tool_result immediately following a non-assistant message (i.e. an
+        // orphaned tool result without a preceding assistant ToolUse).
+        let messages = &result.compacted_session.messages;
+        for i in 1..messages.len() {
+            let curr_is_tool_result = messages[i]
+                .blocks
+                .first()
+                .map(|b| matches!(b, ContentBlock::ToolResult { .. }))
+                .unwrap_or(false);
+            if curr_is_tool_result {
+                let prev_has_tool_use = messages[i - 1]
+                    .blocks
+                    .iter()
+                    .any(|b| matches!(b, ContentBlock::ToolUse { .. }));
+                assert!(
+                    prev_has_tool_use,
+                    "message[{}] is a ToolResult but message[{}] has no ToolUse: {:?}",
+                    i,
+                    i - 1,
+                    &messages[i - 1].blocks
+                );
+            }
+        }
    }

    #[test]
@@ -0,0 +1,901 @@
+use std::collections::BTreeMap;
+use std::path::Path;
+
+use crate::config::ConfigError;
+use crate::json::JsonValue;
+
+/// Diagnostic emitted when a config file contains a suspect field.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ConfigDiagnostic {
+    pub path: String,
+    pub field: String,
+    pub line: Option<usize>,
+    pub kind: DiagnosticKind,
+}
+
+/// Classification of the diagnostic.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum DiagnosticKind {
+    UnknownKey {
+        suggestion: Option<String>,
+    },
+    WrongType {
+        expected: &'static str,
+        got: &'static str,
+    },
+    Deprecated {
+        replacement: &'static str,
+    },
+}
+
+impl std::fmt::Display for ConfigDiagnostic {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        let location = self
+            .line
+            .map_or_else(String::new, |line| format!(" (line {line})"));
+        match &self.kind {
+            DiagnosticKind::UnknownKey { suggestion: None } => {
+                write!(f, "{}: unknown key \"{}\"{location}", self.path, self.field)
+            }
+            DiagnosticKind::UnknownKey {
+                suggestion: Some(hint),
+            } => {
+                write!(
+                    f,
+                    "{}: unknown key \"{}\"{location}. Did you mean \"{}\"?",
+                    self.path, self.field, hint
+                )
+            }
+            DiagnosticKind::WrongType { expected, got } => {
+                write!(
+                    f,
+                    "{}: field \"{}\" must be {expected}, got {got}{location}",
+                    self.path, self.field
+                )
+            }
+            DiagnosticKind::Deprecated { replacement } => {
+                write!(
+                    f,
+                    "{}: field \"{}\" is deprecated{location}. Use \"{replacement}\" instead",
+                    self.path, self.field
+                )
+            }
+        }
+    }
+}
+
+/// Result of validating a single config file.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ValidationResult {
+    pub errors: Vec<ConfigDiagnostic>,
+    pub warnings: Vec<ConfigDiagnostic>,
+}
+
+impl ValidationResult {
+    #[must_use]
+    pub fn is_ok(&self) -> bool {
+        self.errors.is_empty()
+    }
+
+    fn merge(&mut self, other: Self) {
+        self.errors.extend(other.errors);
+        self.warnings.extend(other.warnings);
+    }
+}
+
+// ---- known-key schema ----
+
+/// Expected type for a config field.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum FieldType {
+    String,
+    Bool,
+    Object,
+    StringArray,
+    Number,
+}
+
+impl FieldType {
+    fn label(self) -> &'static str {
+        match self {
+            Self::String => "a string",
+            Self::Bool => "a boolean",
+            Self::Object => "an object",
+            Self::StringArray => "an array of strings",
+            Self::Number => "a number",
+        }
+    }
+
+    fn matches(self, value: &JsonValue) -> bool {
+        match self {
+            Self::String => value.as_str().is_some(),
+            Self::Bool => value.as_bool().is_some(),
+            Self::Object => value.as_object().is_some(),
+            Self::StringArray => value
+                .as_array()
+                .is_some_and(|arr| arr.iter().all(|v| v.as_str().is_some())),
+            Self::Number => value.as_i64().is_some(),
+        }
+    }
+}
+
+fn json_type_label(value: &JsonValue) -> &'static str {
+    match value {
+        JsonValue::Null => "null",
+        JsonValue::Bool(_) => "a boolean",
+        JsonValue::Number(_) => "a number",
+        JsonValue::String(_) => "a string",
+        JsonValue::Array(_) => "an array",
+        JsonValue::Object(_) => "an object",
+    }
+}
+
+struct FieldSpec {
+    name: &'static str,
+    expected: FieldType,
+}
+
+struct DeprecatedField {
+    name: &'static str,
+    replacement: &'static str,
+}
+
+const TOP_LEVEL_FIELDS: &[FieldSpec] = &[
+    FieldSpec {
+        name: "$schema",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "model",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "hooks",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "permissions",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "permissionMode",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "mcpServers",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "oauth",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "enabledPlugins",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "plugins",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "sandbox",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "env",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "aliases",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "providerFallbacks",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "trustedRoots",
+        expected: FieldType::StringArray,
+    },
+];
+
+const HOOKS_FIELDS: &[FieldSpec] = &[
+    FieldSpec {
+        name: "PreToolUse",
+        expected: FieldType::StringArray,
+    },
+    FieldSpec {
+        name: "PostToolUse",
+        expected: FieldType::StringArray,
+    },
+    FieldSpec {
+        name: "PostToolUseFailure",
+        expected: FieldType::StringArray,
+    },
+];
+
+const PERMISSIONS_FIELDS: &[FieldSpec] = &[
+    FieldSpec {
+        name: "defaultMode",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "allow",
+        expected: FieldType::StringArray,
+    },
+    FieldSpec {
+        name: "deny",
+        expected: FieldType::StringArray,
+    },
+    FieldSpec {
+        name: "ask",
+        expected: FieldType::StringArray,
+    },
+];
+
+const PLUGINS_FIELDS: &[FieldSpec] = &[
+    FieldSpec {
+        name: "enabled",
+        expected: FieldType::Object,
+    },
+    FieldSpec {
+        name: "externalDirectories",
+        expected: FieldType::StringArray,
+    },
+    FieldSpec {
+        name: "installRoot",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "registryPath",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "bundledRoot",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "maxOutputTokens",
+        expected: FieldType::Number,
+    },
+];
+
+const SANDBOX_FIELDS: &[FieldSpec] = &[
+    FieldSpec {
+        name: "enabled",
+        expected: FieldType::Bool,
+    },
+    FieldSpec {
+        name: "namespaceRestrictions",
+        expected: FieldType::Bool,
+    },
+    FieldSpec {
+        name: "networkIsolation",
+        expected: FieldType::Bool,
+    },
+    FieldSpec {
+        name: "filesystemMode",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "allowedMounts",
+        expected: FieldType::StringArray,
+    },
+];
+
+const OAUTH_FIELDS: &[FieldSpec] = &[
+    FieldSpec {
+        name: "clientId",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "authorizeUrl",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "tokenUrl",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "callbackPort",
+        expected: FieldType::Number,
+    },
+    FieldSpec {
+        name: "manualRedirectUrl",
+        expected: FieldType::String,
+    },
+    FieldSpec {
+        name: "scopes",
+        expected: FieldType::StringArray,
+    },
+];
+
+const DEPRECATED_FIELDS: &[DeprecatedField] = &[
+    DeprecatedField {
+        name: "permissionMode",
+        replacement: "permissions.defaultMode",
+    },
+    DeprecatedField {
+        name: "enabledPlugins",
+        replacement: "plugins.enabled",
+    },
+];
+
+// ---- line-number resolution ----
+
+/// Find the 1-based line number where a JSON key first appears in the raw source.
+fn find_key_line(source: &str, key: &str) -> Option<usize> {
+    // Search for `"key"` followed by optional whitespace and a colon.
+    let needle = format!("\"{key}\"");
+    let mut search_start = 0;
+    while let Some(offset) = source[search_start..].find(&needle) {
+        let absolute = search_start + offset;
+        let after = absolute + needle.len();
+        // Verify the next non-whitespace char is `:` to confirm this is a key, not a value.
+        if source[after..].chars().find(|ch| !ch.is_ascii_whitespace()) == Some(':') {
+            return Some(source[..absolute].chars().filter(|&ch| ch == '\n').count() + 1);
+        }
+        search_start = after;
+    }
+    None
+}
+
+// ---- core validation ----
+
+fn validate_object_keys(
+    object: &BTreeMap<String, JsonValue>,
+    known_fields: &[FieldSpec],
+    prefix: &str,
+    source: &str,
+    path_display: &str,
+) -> ValidationResult {
+    let mut result = ValidationResult {
+        errors: Vec::new(),
+        warnings: Vec::new(),
+    };
+
+    let known_names: Vec<&str> = known_fields.iter().map(|f| f.name).collect();
+
+    for (key, value) in object {
+        let field_path = if prefix.is_empty() {
+            key.clone()
+        } else {
+            format!("{prefix}.{key}")
+        };
+
+        if let Some(spec) = known_fields.iter().find(|f| f.name == key) {
+            // Type check.
+            if !spec.expected.matches(value) {
+                result.errors.push(ConfigDiagnostic {
+                    path: path_display.to_string(),
+                    field: field_path,
+                    line: find_key_line(source, key),
+                    kind: DiagnosticKind::WrongType {
+                        expected: spec.expected.label(),
+                        got: json_type_label(value),
+                    },
+                });
+            }
+        } else if DEPRECATED_FIELDS.iter().any(|d| d.name == key) {
+            // Deprecated key — handled separately, not an unknown-key error.
+        } else {
+            // Unknown key.
+            let suggestion = suggest_field(key, &known_names);
+            result.errors.push(ConfigDiagnostic {
+                path: path_display.to_string(),
+                field: field_path,
+                line: find_key_line(source, key),
+                kind: DiagnosticKind::UnknownKey { suggestion },
+            });
+        }
+    }
+
+    result
+}
+
+fn suggest_field(input: &str, candidates: &[&str]) -> Option<String> {
+    let input_lower = input.to_ascii_lowercase();
+    candidates
+        .iter()
+        .filter_map(|candidate| {
+            let distance = simple_edit_distance(&input_lower, &candidate.to_ascii_lowercase());
+            (distance <= 3).then_some((distance, *candidate))
+        })
+        .min_by_key(|(distance, _)| *distance)
+        .map(|(_, name)| name.to_string())
+}
+
+fn simple_edit_distance(left: &str, right: &str) -> usize {
+    if left.is_empty() {
+        return right.len();
+    }
+    if right.is_empty() {
+        return left.len();
+    }
+    let right_chars: Vec<char> = right.chars().collect();
+    let mut previous: Vec<usize> = (0..=right_chars.len()).collect();
+    let mut current = vec![0; right_chars.len() + 1];
+
+    for (left_index, left_char) in left.chars().enumerate() {
+        current[0] = left_index + 1;
+        for (right_index, right_char) in right_chars.iter().enumerate() {
+            let cost = usize::from(left_char != *right_char);
+            current[right_index + 1] = (previous[right_index + 1] + 1)
+                .min(current[right_index] + 1)
+                .min(previous[right_index] + cost);
+        }
+        previous.clone_from(&current);
+    }
+
+    previous[right_chars.len()]
+}
+
+/// Validate a parsed config file's keys and types against the known schema.
+///
+/// Returns diagnostics (errors and deprecation warnings) without blocking the load.
+pub fn validate_config_file(
+    object: &BTreeMap<String, JsonValue>,
+    source: &str,
+    file_path: &Path,
+) -> ValidationResult {
+    let path_display = file_path.display().to_string();
+    let mut result = validate_object_keys(object, TOP_LEVEL_FIELDS, "", source, &path_display);
+
+    // Check deprecated fields.
+    for deprecated in DEPRECATED_FIELDS {
+        if object.contains_key(deprecated.name) {
+            result.warnings.push(ConfigDiagnostic {
+                path: path_display.clone(),
+                field: deprecated.name.to_string(),
+                line: find_key_line(source, deprecated.name),
+                kind: DiagnosticKind::Deprecated {
+                    replacement: deprecated.replacement,
+                },
+            });
+        }
+    }
+
+    // Validate known nested objects.
+    if let Some(hooks) = object.get("hooks").and_then(JsonValue::as_object) {
+        result.merge(validate_object_keys(
+            hooks,
+            HOOKS_FIELDS,
+            "hooks",
+            source,
+            &path_display,
+        ));
+    }
+    if let Some(permissions) = object.get("permissions").and_then(JsonValue::as_object) {
+        result.merge(validate_object_keys(
+            permissions,
+            PERMISSIONS_FIELDS,
+            "permissions",
+            source,
+            &path_display,
+        ));
+    }
+    if let Some(plugins) = object.get("plugins").and_then(JsonValue::as_object) {
+        result.merge(validate_object_keys(
+            plugins,
+            PLUGINS_FIELDS,
+            "plugins",
+            source,
+            &path_display,
+        ));
+    }
+    if let Some(sandbox) = object.get("sandbox").and_then(JsonValue::as_object) {
+        result.merge(validate_object_keys(
+            sandbox,
+            SANDBOX_FIELDS,
+            "sandbox",
+            source,
+            &path_display,
+        ));
+    }
+    if let Some(oauth) = object.get("oauth").and_then(JsonValue::as_object) {
+        result.merge(validate_object_keys(
+            oauth,
+            OAUTH_FIELDS,
+            "oauth",
+            source,
+            &path_display,
+        ));
+    }
+
+    result
+}
+
+/// Check whether a file path uses an unsupported config format (e.g. TOML).
+pub fn check_unsupported_format(file_path: &Path) -> Result<(), ConfigError> {
+    if let Some(ext) = file_path.extension().and_then(|e| e.to_str()) {
+        if ext.eq_ignore_ascii_case("toml") {
+            return Err(ConfigError::Parse(format!(
+                "{}: TOML config files are not supported. Use JSON (settings.json) instead",
+                file_path.display()
+            )));
+        }
+    }
+    Ok(())
+}
+
+/// Format all diagnostics into a human-readable report.
+#[must_use]
+pub fn format_diagnostics(result: &ValidationResult) -> String {
+    let mut lines = Vec::new();
+    for warning in &result.warnings {
+        lines.push(format!("warning: {warning}"));
+    }
+    for error in &result.errors {
+        lines.push(format!("error: {error}"));
+    }
+    lines.join("\n")
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::path::PathBuf;
+
+    fn test_path() -> PathBuf {
+        PathBuf::from("/test/settings.json")
+    }
+
+    #[test]
+    fn detects_unknown_top_level_key() {
+        // given
+        let source = r#"{"model": "opus", "unknownField": true}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "unknownField");
+        assert!(matches!(
+            result.errors[0].kind,
+            DiagnosticKind::UnknownKey { .. }
+        ));
+    }
+
+    #[test]
+    fn detects_wrong_type_for_model() {
+        // given
+        let source = r#"{"model": 123}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "model");
+        assert!(matches!(
+            result.errors[0].kind,
+            DiagnosticKind::WrongType {
+                expected: "a string",
+                got: "a number"
+            }
+        ));
+    }
+
+    #[test]
+    fn detects_deprecated_permission_mode() {
+        // given
+        let source = r#"{"permissionMode": "plan"}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.warnings.len(), 1);
+        assert_eq!(result.warnings[0].field, "permissionMode");
+        assert!(matches!(
+            result.warnings[0].kind,
+            DiagnosticKind::Deprecated {
+                replacement: "permissions.defaultMode"
+            }
+        ));
+    }
+
+    #[test]
+    fn detects_deprecated_enabled_plugins() {
+        // given
+        let source = r#"{"enabledPlugins": {"tool-guard@builtin": true}}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.warnings.len(), 1);
+        assert_eq!(result.warnings[0].field, "enabledPlugins");
+        assert!(matches!(
+            result.warnings[0].kind,
+            DiagnosticKind::Deprecated {
+                replacement: "plugins.enabled"
+            }
+        ));
+    }
+
+    #[test]
+    fn reports_line_number_for_unknown_key() {
+        // given
+        let source = "{\n  \"model\": \"opus\",\n  \"badKey\": true\n}";
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].line, Some(3));
+        assert_eq!(result.errors[0].field, "badKey");
+    }
+
+    #[test]
+    fn reports_line_number_for_wrong_type() {
+        // given
+        let source = "{\n  \"model\": 42\n}";
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].line, Some(2));
+    }
+
+    #[test]
+    fn validates_nested_hooks_keys() {
+        // given
+        let source = r#"{"hooks": {"PreToolUse": ["cmd"], "BadHook": ["x"]}}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "hooks.BadHook");
+    }
+
+    #[test]
+    fn validates_nested_permissions_keys() {
+        // given
+        let source = r#"{"permissions": {"allow": ["Read"], "denyAll": true}}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "permissions.denyAll");
+    }
+
+    #[test]
+    fn validates_nested_sandbox_keys() {
+        // given
+        let source = r#"{"sandbox": {"enabled": true, "containerMode": "strict"}}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "sandbox.containerMode");
+    }
+
+    #[test]
+    fn validates_nested_plugins_keys() {
+        // given
+        let source = r#"{"plugins": {"installRoot": "/tmp", "autoUpdate": true}}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "plugins.autoUpdate");
+    }
+
+    #[test]
+    fn validates_nested_oauth_keys() {
+        // given
+        let source = r#"{"oauth": {"clientId": "abc", "secret": "hidden"}}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "oauth.secret");
+    }
+
+    #[test]
+    fn valid_config_produces_no_diagnostics() {
+        // given
+        let source = r#"{
+  "model": "opus",
+  "hooks": {"PreToolUse": ["guard"]},
+  "permissions": {"defaultMode": "plan", "allow": ["Read"]},
+  "mcpServers": {},
+  "sandbox": {"enabled": false}
+}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert!(result.is_ok());
+        assert!(result.warnings.is_empty());
+    }
+
+    #[test]
+    fn suggests_close_field_name() {
+        // given
+        let source = r#"{"modle": "opus"}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        match &result.errors[0].kind {
+            DiagnosticKind::UnknownKey {
+                suggestion: Some(s),
+            } => assert_eq!(s, "model"),
+            other => panic!("expected suggestion, got {other:?}"),
+        }
+    }
+
+    #[test]
+    fn format_diagnostics_includes_all_entries() {
+        // given
+        let source = r#"{"permissionMode": "plan", "badKey": 1}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+        let result = validate_config_file(object, source, &test_path());
+
+        // when
+        let output = format_diagnostics(&result);
+
+        // then
+        assert!(output.contains("warning:"));
+        assert!(output.contains("error:"));
+        assert!(output.contains("badKey"));
+        assert!(output.contains("permissionMode"));
+    }
+
+    #[test]
+    fn check_unsupported_format_rejects_toml() {
+        // given
+        let path = PathBuf::from("/home/.claw/settings.toml");
+
+        // when
+        let result = check_unsupported_format(&path);
+
+        // then
+        assert!(result.is_err());
+        let message = result.unwrap_err().to_string();
+        assert!(message.contains("TOML"));
+        assert!(message.contains("settings.toml"));
+    }
+
+    #[test]
+    fn check_unsupported_format_allows_json() {
+        // given
+        let path = PathBuf::from("/home/.claw/settings.json");
+
+        // when / then
+        assert!(check_unsupported_format(&path).is_ok());
+    }
+
+    #[test]
+    fn wrong_type_in_nested_sandbox_field() {
+        // given
+        let source = r#"{"sandbox": {"enabled": "yes"}}"#;
+        let parsed = JsonValue::parse(source).expect("valid json");
+        let object = parsed.as_object().expect("object");
+
+        // when
+        let result = validate_config_file(object, source, &test_path());
+
+        // then
+        assert_eq!(result.errors.len(), 1);
+        assert_eq!(result.errors[0].field, "sandbox.enabled");
+        assert!(matches!(
+            result.errors[0].kind,
+            DiagnosticKind::WrongType {
+                expected: "a boolean",
+                got: "a string"
+            }
+        ));
+    }
+
+    #[test]
+    fn display_format_unknown_key_with_line() {
+        // given
+        let diag = ConfigDiagnostic {
+            path: "/test/settings.json".to_string(),
+            field: "badKey".to_string(),
+            line: Some(5),
+            kind: DiagnosticKind::UnknownKey { suggestion: None },
+        };
+
+        // when
+        let output = diag.to_string();
+
+        // then
+        assert_eq!(
+            output,
+            r#"/test/settings.json: unknown key "badKey" (line 5)"#
+        );
+    }
+
+    #[test]
+    fn display_format_wrong_type_with_line() {
+        // given
+        let diag = ConfigDiagnostic {
+            path: "/test/settings.json".to_string(),
+            field: "model".to_string(),
+            line: Some(2),
+            kind: DiagnosticKind::WrongType {
+                expected: "a string",
+                got: "a number",
+            },
+        };
+
+        // when
+        let output = diag.to_string();
+
+        // then
+        assert_eq!(
+            output,
+            r#"/test/settings.json: field "model" must be a string, got a number (line 2)"#
+        );
+    }
+
+    #[test]
+    fn display_format_deprecated_with_line() {
+        // given
+        let diag = ConfigDiagnostic {
+            path: "/test/settings.json".to_string(),
+            field: "permissionMode".to_string(),
+            line: Some(3),
+            kind: DiagnosticKind::Deprecated {
+                replacement: "permissions.defaultMode",
+            },
+        };
+
+        // when
+        let output = diag.to_string();
+
+        // then
+        assert_eq!(
+            output,
+            r#"/test/settings.json: field "permissionMode" is deprecated (line 3). Use "permissions.defaultMode" instead"#
+        );
+    }
+}
@@ -9,6 +9,41 @@ use regex::RegexBuilder;
 use serde::{Deserialize, Serialize};
 use walkdir::WalkDir;

+/// Maximum file size that can be read (10 MB).
+const MAX_READ_SIZE: u64 = 10 * 1024 * 1024;
+
+/// Maximum file size that can be written (10 MB).
+const MAX_WRITE_SIZE: usize = 10 * 1024 * 1024;
+
+/// Check whether a file appears to contain binary content by examining
+/// the first chunk for NUL bytes.
+fn is_binary_file(path: &Path) -> io::Result<bool> {
+    use std::io::Read;
+    let mut file = fs::File::open(path)?;
+    let mut buffer = [0u8; 8192];
+    let bytes_read = file.read(&mut buffer)?;
+    Ok(buffer[..bytes_read].contains(&0))
+}
+
+/// Validate that a resolved path stays within the given workspace root.
+/// Returns the canonical path on success, or an error if the path escapes
+/// the workspace boundary (e.g. via `../` traversal or symlink).
+#[allow(dead_code)]
+fn validate_workspace_boundary(resolved: &Path, workspace_root: &Path) -> io::Result<()> {
+    if !resolved.starts_with(workspace_root) {
+        return Err(io::Error::new(
+            io::ErrorKind::PermissionDenied,
+            format!(
+                "path {} escapes workspace boundary {}",
+                resolved.display(),
+                workspace_root.display()
+            ),
+        ));
+    }
+    Ok(())
+}
+
+/// Text payload returned by file-reading operations.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct TextFilePayload {
    #[serde(rename = "filePath")]
@@ -22,6 +57,7 @@ pub struct TextFilePayload {
    pub total_lines: usize,
 }

+/// Output envelope for the `read_file` tool.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct ReadFileOutput {
    #[serde(rename = "type")]
@@ -29,6 +65,7 @@ pub struct ReadFileOutput {
    pub file: TextFilePayload,
 }

+/// Structured patch hunk emitted by write and edit operations.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct StructuredPatchHunk {
    #[serde(rename = "oldStart")]
@@ -42,6 +79,7 @@ pub struct StructuredPatchHunk {
    pub lines: Vec<String>,
 }

+/// Output envelope for full-file write operations.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct WriteFileOutput {
    #[serde(rename = "type")]
@@ -57,6 +95,7 @@ pub struct WriteFileOutput {
    pub git_diff: Option<serde_json::Value>,
 }

+/// Output envelope for targeted string-replacement edits.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct EditFileOutput {
    #[serde(rename = "filePath")]
@@ -77,6 +116,7 @@ pub struct EditFileOutput {
    pub git_diff: Option<serde_json::Value>,
 }

+/// Result of a glob-based filename search.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 pub struct GlobSearchOutput {
    #[serde(rename = "durationMs")]
@@ -87,6 +127,7 @@ pub struct GlobSearchOutput {
    pub truncated: bool,
 }

+/// Parameters accepted by the grep-style search tool.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct GrepSearchInput {
    pub pattern: String,
@@ -112,6 +153,7 @@ pub struct GrepSearchInput {
    pub multiline: Option<bool>,
 }

+/// Result payload returned by the grep-style search tool.
 #[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
 pub struct GrepSearchOutput {
    pub mode: Option<String>,
@@ -129,12 +171,35 @@ pub struct GrepSearchOutput {
    pub applied_offset: Option<usize>,
 }

+/// Reads a text file and returns a line-windowed payload.
 pub fn read_file(
    path: &str,
    offset: Option<usize>,
    limit: Option<usize>,
 ) -> io::Result<ReadFileOutput> {
    let absolute_path = normalize_path(path)?;
+
+    // Check file size before reading
+    let metadata = fs::metadata(&absolute_path)?;
+    if metadata.len() > MAX_READ_SIZE {
+        return Err(io::Error::new(
+            io::ErrorKind::InvalidData,
+            format!(
+                "file is too large ({} bytes, max {} bytes)",
+                metadata.len(),
+                MAX_READ_SIZE
+            ),
+        ));
+    }
+
+    // Detect binary files
+    if is_binary_file(&absolute_path)? {
+        return Err(io::Error::new(
+            io::ErrorKind::InvalidData,
+            "file appears to be binary",
+        ));
+    }
+
    let content = fs::read_to_string(&absolute_path)?;
    let lines: Vec<&str> = content.lines().collect();
    let start_index = offset.unwrap_or(0).min(lines.len());
@@ -155,7 +220,19 @@ pub fn read_file(
    })
 }

+/// Replaces a file's contents and returns patch metadata.
 pub fn write_file(path: &str, content: &str) -> io::Result<WriteFileOutput> {
+    if content.len() > MAX_WRITE_SIZE {
+        return Err(io::Error::new(
+            io::ErrorKind::InvalidData,
+            format!(
+                "content is too large ({} bytes, max {} bytes)",
+                content.len(),
+                MAX_WRITE_SIZE
+            ),
+        ));
+    }
+
    let absolute_path = normalize_path_allow_missing(path)?;
    let original_file = fs::read_to_string(&absolute_path).ok();
    if let Some(parent) = absolute_path.parent() {
@@ -177,6 +254,7 @@ pub fn write_file(path: &str, content: &str) -> io::Result<WriteFileOutput> {
    })
 }

+/// Performs an in-file string replacement and returns patch metadata.
 pub fn edit_file(
    path: &str,
    old_string: &str,
@@ -217,6 +295,7 @@ pub fn edit_file(
    })
 }

+/// Expands a glob pattern and returns matching filenames.
 pub fn glob_search(pattern: &str, path: Option<&str>) -> io::Result<GlobSearchOutput> {
    let started = Instant::now();
    let base_dir = path
@@ -229,12 +308,20 @@ pub fn glob_search(pattern: &str, path: Option<&str>) -> io::Result<GlobSearchOu
        base_dir.join(pattern).to_string_lossy().into_owned()
    };

+    // The `glob` crate does not support brace expansion ({a,b,c}).
+    // Expand braces into multiple patterns so patterns like
+    // `Assets/**/*.{cs,uxml,uss}` work correctly.
+    let expanded = expand_braces(&search_pattern);
+
+    let mut seen = std::collections::HashSet::new();
    let mut matches = Vec::new();
-    let entries = glob::glob(&search_pattern)
-        .map_err(|error| io::Error::new(io::ErrorKind::InvalidInput, error.to_string()))?;
-    for entry in entries.flatten() {
-        if entry.is_file() {
-            matches.push(entry);
+    for pat in &expanded {
+        let entries = glob::glob(pat)
+            .map_err(|error| io::Error::new(io::ErrorKind::InvalidInput, error.to_string()))?;
+        for entry in entries.flatten() {
+            if entry.is_file() && seen.insert(entry.clone()) {
+                matches.push(entry);
+            }
        }
    }

@@ -260,6 +347,7 @@ pub fn glob_search(pattern: &str, path: Option<&str>) -> io::Result<GlobSearchOu
    })
 }

+/// Runs a regex search over workspace files with optional context lines.
 pub fn grep_search(input: &GrepSearchInput) -> io::Result<GrepSearchOutput> {
    let base_path = input
        .path
@@ -477,18 +565,105 @@ fn normalize_path_allow_missing(path: &str) -> io::Result<PathBuf> {
    Ok(candidate)
 }

+/// Read a file with workspace boundary enforcement.
+#[allow(dead_code)]
+pub fn read_file_in_workspace(
+    path: &str,
+    offset: Option<usize>,
+    limit: Option<usize>,
+    workspace_root: &Path,
+) -> io::Result<ReadFileOutput> {
+    let absolute_path = normalize_path(path)?;
+    let canonical_root = workspace_root
+        .canonicalize()
+        .unwrap_or_else(|_| workspace_root.to_path_buf());
+    validate_workspace_boundary(&absolute_path, &canonical_root)?;
+    read_file(path, offset, limit)
+}
+
+/// Write a file with workspace boundary enforcement.
+#[allow(dead_code)]
+pub fn write_file_in_workspace(
+    path: &str,
+    content: &str,
+    workspace_root: &Path,
+) -> io::Result<WriteFileOutput> {
+    let absolute_path = normalize_path_allow_missing(path)?;
+    let canonical_root = workspace_root
+        .canonicalize()
+        .unwrap_or_else(|_| workspace_root.to_path_buf());
+    validate_workspace_boundary(&absolute_path, &canonical_root)?;
+    write_file(path, content)
+}
+
+/// Edit a file with workspace boundary enforcement.
+#[allow(dead_code)]
+pub fn edit_file_in_workspace(
+    path: &str,
+    old_string: &str,
+    new_string: &str,
+    replace_all: bool,
+    workspace_root: &Path,
+) -> io::Result<EditFileOutput> {
+    let absolute_path = normalize_path(path)?;
+    let canonical_root = workspace_root
+        .canonicalize()
+        .unwrap_or_else(|_| workspace_root.to_path_buf());
+    validate_workspace_boundary(&absolute_path, &canonical_root)?;
+    edit_file(path, old_string, new_string, replace_all)
+}
+
+/// Check whether a path is a symlink that resolves outside the workspace.
+#[allow(dead_code)]
+pub fn is_symlink_escape(path: &Path, workspace_root: &Path) -> io::Result<bool> {
+    let metadata = fs::symlink_metadata(path)?;
+    if !metadata.is_symlink() {
+        return Ok(false);
+    }
+    let resolved = path.canonicalize()?;
+    let canonical_root = workspace_root
+        .canonicalize()
+        .unwrap_or_else(|_| workspace_root.to_path_buf());
+    Ok(!resolved.starts_with(&canonical_root))
+}
+
+/// Expand shell-style brace groups in a glob pattern.
+///
+/// Handles one level of braces: `foo.{a,b,c}` → `["foo.a", "foo.b", "foo.c"]`.
+/// Nested braces are not expanded (uncommon in practice).
+/// Patterns without braces pass through unchanged.
+fn expand_braces(pattern: &str) -> Vec<String> {
+    let Some(open) = pattern.find('{') else {
+        return vec![pattern.to_owned()];
+    };
+    let Some(close) = pattern[open..].find('}').map(|i| open + i) else {
+        // Unmatched brace — treat as literal.
+        return vec![pattern.to_owned()];
+    };
+    let prefix = &pattern[..open];
+    let suffix = &pattern[close + 1..];
+    let alternatives = &pattern[open + 1..close];
+    alternatives
+        .split(',')
+        .flat_map(|alt| expand_braces(&format!("{prefix}{alt}{suffix}")))
+        .collect()
+}
+
 #[cfg(test)]
 mod tests {
    use std::time::{SystemTime, UNIX_EPOCH};

-    use super::{edit_file, glob_search, grep_search, read_file, write_file, GrepSearchInput};
+    use super::{
+        edit_file, expand_braces, glob_search, grep_search, is_symlink_escape, read_file,
+        read_file_in_workspace, write_file, GrepSearchInput, MAX_WRITE_SIZE,
+    };

    fn temp_path(name: &str) -> std::path::PathBuf {
        let unique = SystemTime::now()
            .duration_since(UNIX_EPOCH)
            .expect("time should move forward")
            .as_nanos();
-        std::env::temp_dir().join(format!("claw-native-{name}-{unique}"))
+        std::env::temp_dir().join(format!("clawd-native-{name}-{unique}"))
    }

    #[test]
@@ -513,6 +688,73 @@ mod tests {
        assert!(output.replace_all);
    }

+    #[test]
+    fn rejects_binary_files() {
+        let path = temp_path("binary-test.bin");
+        std::fs::write(&path, b"\x00\x01\x02\x03binary content").expect("write should succeed");
+        let result = read_file(path.to_string_lossy().as_ref(), None, None);
+        assert!(result.is_err());
+        let error = result.unwrap_err();
+        assert_eq!(error.kind(), std::io::ErrorKind::InvalidData);
+        assert!(error.to_string().contains("binary"));
+    }
+
+    #[test]
+    fn rejects_oversized_writes() {
+        let path = temp_path("oversize-write.txt");
+        let huge = "x".repeat(MAX_WRITE_SIZE + 1);
+        let result = write_file(path.to_string_lossy().as_ref(), &huge);
+        assert!(result.is_err());
+        let error = result.unwrap_err();
+        assert_eq!(error.kind(), std::io::ErrorKind::InvalidData);
+        assert!(error.to_string().contains("too large"));
+    }
+
+    #[test]
+    fn enforces_workspace_boundary() {
+        let workspace = temp_path("workspace-boundary");
+        std::fs::create_dir_all(&workspace).expect("workspace dir should be created");
+        let inside = workspace.join("inside.txt");
+        write_file(inside.to_string_lossy().as_ref(), "safe content")
+            .expect("write inside workspace should succeed");
+
+        // Reading inside workspace should succeed
+        let result =
+            read_file_in_workspace(inside.to_string_lossy().as_ref(), None, None, &workspace);
+        assert!(result.is_ok());
+
+        // Reading outside workspace should fail
+        let outside = temp_path("outside-boundary.txt");
+        write_file(outside.to_string_lossy().as_ref(), "unsafe content")
+            .expect("write outside should succeed");
+        let result =
+            read_file_in_workspace(outside.to_string_lossy().as_ref(), None, None, &workspace);
+        assert!(result.is_err());
+        let error = result.unwrap_err();
+        assert_eq!(error.kind(), std::io::ErrorKind::PermissionDenied);
+        assert!(error.to_string().contains("escapes workspace"));
+    }
+
+    #[test]
+    fn detects_symlink_escape() {
+        let workspace = temp_path("symlink-workspace");
+        std::fs::create_dir_all(&workspace).expect("workspace dir should be created");
+        let outside = temp_path("symlink-target.txt");
+        std::fs::write(&outside, "target content").expect("target should write");
+
+        let _link_path = workspace.join("escape-link.txt");
+        #[cfg(unix)]
+        {
+            std::os::unix::fs::symlink(&outside, &link_path).expect("symlink should create");
+            assert!(is_symlink_escape(&link_path, &workspace).expect("check should succeed"));
+        }
+
+        // Non-symlink file should not be an escape
+        let normal = workspace.join("normal.txt");
+        std::fs::write(&normal, "normal content").expect("normal file should write");
+        assert!(!is_symlink_escape(&normal, &workspace).expect("check should succeed"));
+    }
+
    #[test]
    fn globs_and_greps_directory() {
        let dir = temp_path("search-dir");
@@ -547,4 +789,51 @@ mod tests {
        .expect("grep should succeed");
        assert!(grep_output.content.unwrap_or_default().contains("hello"));
    }
+
+    #[test]
+    fn expand_braces_no_braces() {
+        assert_eq!(expand_braces("*.rs"), vec!["*.rs"]);
+    }
+
+    #[test]
+    fn expand_braces_single_group() {
+        let mut result = expand_braces("Assets/**/*.{cs,uxml,uss}");
+        result.sort();
+        assert_eq!(
+            result,
+            vec!["Assets/**/*.cs", "Assets/**/*.uss", "Assets/**/*.uxml",]
+        );
+    }
+
+    #[test]
+    fn expand_braces_nested() {
+        let mut result = expand_braces("src/{a,b}.{rs,toml}");
+        result.sort();
+        assert_eq!(
+            result,
+            vec!["src/a.rs", "src/a.toml", "src/b.rs", "src/b.toml"]
+        );
+    }
+
+    #[test]
+    fn expand_braces_unmatched() {
+        assert_eq!(expand_braces("foo.{bar"), vec!["foo.{bar"]);
+    }
+
+    #[test]
+    fn glob_search_with_braces_finds_files() {
+        let dir = temp_path("glob-braces");
+        std::fs::create_dir_all(&dir).unwrap();
+        std::fs::write(dir.join("a.rs"), "fn main() {}").unwrap();
+        std::fs::write(dir.join("b.toml"), "[package]").unwrap();
+        std::fs::write(dir.join("c.txt"), "hello").unwrap();
+
+        let result =
+            glob_search("*.{rs,toml}", Some(dir.to_str().unwrap())).expect("glob should succeed");
+        assert_eq!(
+            result.num_files, 2,
+            "should match .rs and .toml but not .txt"
+        );
+        let _ = std::fs::remove_dir_all(&dir);
+    }
 }
@@ -0,0 +1,324 @@
+use std::path::Path;
+use std::process::Command;
+
+/// A single git commit entry from the log.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct GitCommitEntry {
+    pub hash: String,
+    pub subject: String,
+}
+
+/// Git-aware context gathered at startup for injection into the system prompt.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct GitContext {
+    pub branch: Option<String>,
+    pub recent_commits: Vec<GitCommitEntry>,
+    pub staged_files: Vec<String>,
+}
+
+const MAX_RECENT_COMMITS: usize = 5;
+
+impl GitContext {
+    /// Detect the git context from the given working directory.
+    ///
+    /// Returns `None` when the directory is not inside a git repository.
+    #[must_use]
+    pub fn detect(cwd: &Path) -> Option<Self> {
+        // Quick gate: is this a git repo at all?
+        let rev_parse = Command::new("git")
+            .args(["rev-parse", "--is-inside-work-tree"])
+            .current_dir(cwd)
+            .output()
+            .ok()?;
+        if !rev_parse.status.success() {
+            return None;
+        }
+
+        Some(Self {
+            branch: read_branch(cwd),
+            recent_commits: read_recent_commits(cwd),
+            staged_files: read_staged_files(cwd),
+        })
+    }
+
+    /// Render a human-readable summary suitable for system-prompt injection.
+    #[must_use]
+    pub fn render(&self) -> String {
+        let mut lines = Vec::new();
+
+        if let Some(branch) = &self.branch {
+            lines.push(format!("Git branch: {branch}"));
+        }
+
+        if !self.recent_commits.is_empty() {
+            lines.push(String::new());
+            lines.push("Recent commits:".to_string());
+            for entry in &self.recent_commits {
+                lines.push(format!("  {} {}", entry.hash, entry.subject));
+            }
+        }
+
+        if !self.staged_files.is_empty() {
+            lines.push(String::new());
+            lines.push("Staged files:".to_string());
+            for file in &self.staged_files {
+                lines.push(format!("  {file}"));
+            }
+        }
+
+        lines.join("\n")
+    }
+}
+
+fn read_branch(cwd: &Path) -> Option<String> {
+    let output = Command::new("git")
+        .args(["rev-parse", "--abbrev-ref", "HEAD"])
+        .current_dir(cwd)
+        .output()
+        .ok()?;
+    if !output.status.success() {
+        return None;
+    }
+    let branch = String::from_utf8(output.stdout).ok()?;
+    let trimmed = branch.trim();
+    if trimmed.is_empty() || trimmed == "HEAD" {
+        None
+    } else {
+        Some(trimmed.to_string())
+    }
+}
+
+fn read_recent_commits(cwd: &Path) -> Vec<GitCommitEntry> {
+    let output = Command::new("git")
+        .args([
+            "--no-optional-locks",
+            "log",
+            "--oneline",
+            "-n",
+            &MAX_RECENT_COMMITS.to_string(),
+            "--no-decorate",
+        ])
+        .current_dir(cwd)
+        .output()
+        .ok();
+    let Some(output) = output else {
+        return Vec::new();
+    };
+    if !output.status.success() {
+        return Vec::new();
+    }
+    let stdout = String::from_utf8(output.stdout).unwrap_or_default();
+    stdout
+        .lines()
+        .filter_map(|line| {
+            let line = line.trim();
+            if line.is_empty() {
+                return None;
+            }
+            let (hash, subject) = line.split_once(' ')?;
+            Some(GitCommitEntry {
+                hash: hash.to_string(),
+                subject: subject.to_string(),
+            })
+        })
+        .collect()
+}
+
+fn read_staged_files(cwd: &Path) -> Vec<String> {
+    let output = Command::new("git")
+        .args(["--no-optional-locks", "diff", "--cached", "--name-only"])
+        .current_dir(cwd)
+        .output()
+        .ok();
+    let Some(output) = output else {
+        return Vec::new();
+    };
+    if !output.status.success() {
+        return Vec::new();
+    }
+    let stdout = String::from_utf8(output.stdout).unwrap_or_default();
+    stdout
+        .lines()
+        .filter(|line| !line.trim().is_empty())
+        .map(|line| line.trim().to_string())
+        .collect()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{GitCommitEntry, GitContext};
+    use std::fs;
+    use std::process::Command;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn temp_dir(label: &str) -> std::path::PathBuf {
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_nanos();
+        std::env::temp_dir().join(format!("runtime-git-context-{label}-{nanos}"))
+    }
+
+    fn env_lock() -> std::sync::MutexGuard<'static, ()> {
+        crate::test_env_lock()
+    }
+
+    fn ensure_valid_cwd() {
+        if std::env::current_dir().is_err() {
+            std::env::set_current_dir(env!("CARGO_MANIFEST_DIR"))
+                .expect("test cwd should be recoverable");
+        }
+    }
+
+    #[test]
+    fn returns_none_for_non_git_directory() {
+        // given
+        let _guard = env_lock();
+        ensure_valid_cwd();
+        let root = temp_dir("non-git");
+        fs::create_dir_all(&root).expect("create dir");
+
+        // when
+        let context = GitContext::detect(&root);
+
+        // then
+        assert!(context.is_none());
+        fs::remove_dir_all(root).expect("cleanup");
+    }
+
+    #[test]
+    fn detects_branch_name_and_commits() {
+        // given
+        let _guard = env_lock();
+        ensure_valid_cwd();
+        let root = temp_dir("branch-commits");
+        fs::create_dir_all(&root).expect("create dir");
+        git(&root, &["init", "--quiet", "--initial-branch=main"]);
+        git(&root, &["config", "user.email", "tests@example.com"]);
+        git(&root, &["config", "user.name", "Git Context Tests"]);
+        fs::write(root.join("a.txt"), "a\n").expect("write a");
+        git(&root, &["add", "a.txt"]);
+        git(&root, &["commit", "-m", "first commit", "--quiet"]);
+        fs::write(root.join("b.txt"), "b\n").expect("write b");
+        git(&root, &["add", "b.txt"]);
+        git(&root, &["commit", "-m", "second commit", "--quiet"]);
+
+        // when
+        let context = GitContext::detect(&root).expect("should detect git repo");
+
+        // then
+        assert_eq!(context.branch.as_deref(), Some("main"));
+        assert_eq!(context.recent_commits.len(), 2);
+        assert_eq!(context.recent_commits[0].subject, "second commit");
+        assert_eq!(context.recent_commits[1].subject, "first commit");
+        assert!(context.staged_files.is_empty());
+        fs::remove_dir_all(root).expect("cleanup");
+    }
+
+    #[test]
+    fn detects_staged_files() {
+        // given
+        let _guard = env_lock();
+        ensure_valid_cwd();
+        let root = temp_dir("staged");
+        fs::create_dir_all(&root).expect("create dir");
+        git(&root, &["init", "--quiet", "--initial-branch=main"]);
+        git(&root, &["config", "user.email", "tests@example.com"]);
+        git(&root, &["config", "user.name", "Git Context Tests"]);
+        fs::write(root.join("init.txt"), "init\n").expect("write init");
+        git(&root, &["add", "init.txt"]);
+        git(&root, &["commit", "-m", "initial", "--quiet"]);
+        fs::write(root.join("staged.txt"), "staged\n").expect("write staged");
+        git(&root, &["add", "staged.txt"]);
+
+        // when
+        let context = GitContext::detect(&root).expect("should detect git repo");
+
+        // then
+        assert_eq!(context.staged_files, vec!["staged.txt"]);
+        fs::remove_dir_all(root).expect("cleanup");
+    }
+
+    #[test]
+    fn render_formats_all_sections() {
+        // given
+        let context = GitContext {
+            branch: Some("feat/test".to_string()),
+            recent_commits: vec![
+                GitCommitEntry {
+                    hash: "abc1234".to_string(),
+                    subject: "add feature".to_string(),
+                },
+                GitCommitEntry {
+                    hash: "def5678".to_string(),
+                    subject: "fix bug".to_string(),
+                },
+            ],
+            staged_files: vec!["src/main.rs".to_string()],
+        };
+
+        // when
+        let rendered = context.render();
+
+        // then
+        assert!(rendered.contains("Git branch: feat/test"));
+        assert!(rendered.contains("abc1234 add feature"));
+        assert!(rendered.contains("def5678 fix bug"));
+        assert!(rendered.contains("src/main.rs"));
+    }
+
+    #[test]
+    fn render_omits_empty_sections() {
+        // given
+        let context = GitContext {
+            branch: Some("main".to_string()),
+            recent_commits: Vec::new(),
+            staged_files: Vec::new(),
+        };
+
+        // when
+        let rendered = context.render();
+
+        // then
+        assert!(rendered.contains("Git branch: main"));
+        assert!(!rendered.contains("Recent commits:"));
+        assert!(!rendered.contains("Staged files:"));
+    }
+
+    #[test]
+    fn limits_to_five_recent_commits() {
+        // given
+        let _guard = env_lock();
+        ensure_valid_cwd();
+        let root = temp_dir("five-commits");
+        fs::create_dir_all(&root).expect("create dir");
+        git(&root, &["init", "--quiet", "--initial-branch=main"]);
+        git(&root, &["config", "user.email", "tests@example.com"]);
+        git(&root, &["config", "user.name", "Git Context Tests"]);
+        for i in 1..=8 {
+            let name = format!("file{i}.txt");
+            fs::write(root.join(&name), format!("{i}\n")).expect("write file");
+            git(&root, &["add", &name]);
+            git(&root, &["commit", "-m", &format!("commit {i}"), "--quiet"]);
+        }
+
+        // when
+        let context = GitContext::detect(&root).expect("should detect git repo");
+
+        // then
+        assert_eq!(context.recent_commits.len(), 5);
+        assert_eq!(context.recent_commits[0].subject, "commit 8");
+        assert_eq!(context.recent_commits[4].subject, "commit 4");
+        fs::remove_dir_all(root).expect("cleanup");
+    }
+
+    fn git(cwd: &std::path::Path, args: &[&str]) {
+        let status = Command::new("git")
+            .args(args)
+            .current_dir(cwd)
+            .output()
+            .unwrap_or_else(|_| panic!("git {args:?} should run"))
+            .status;
+        assert!(status.success(), "git {args:?} failed");
+    }
+}
@@ -0,0 +1,152 @@
+use serde::{Deserialize, Serialize};
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum GreenLevel {
+    TargetedTests,
+    Package,
+    Workspace,
+    MergeReady,
+}
+
+impl GreenLevel {
+    #[must_use]
+    pub fn as_str(self) -> &'static str {
+        match self {
+            Self::TargetedTests => "targeted_tests",
+            Self::Package => "package",
+            Self::Workspace => "workspace",
+            Self::MergeReady => "merge_ready",
+        }
+    }
+}
+
+impl std::fmt::Display for GreenLevel {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", self.as_str())
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+pub struct GreenContract {
+    pub required_level: GreenLevel,
+}
+
+impl GreenContract {
+    #[must_use]
+    pub fn new(required_level: GreenLevel) -> Self {
+        Self { required_level }
+    }
+
+    #[must_use]
+    pub fn evaluate(self, observed_level: Option<GreenLevel>) -> GreenContractOutcome {
+        match observed_level {
+            Some(level) if level >= self.required_level => GreenContractOutcome::Satisfied {
+                required_level: self.required_level,
+                observed_level: level,
+            },
+            _ => GreenContractOutcome::Unsatisfied {
+                required_level: self.required_level,
+                observed_level,
+            },
+        }
+    }
+
+    #[must_use]
+    pub fn is_satisfied_by(self, observed_level: GreenLevel) -> bool {
+        observed_level >= self.required_level
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(tag = "outcome", rename_all = "snake_case")]
+pub enum GreenContractOutcome {
+    Satisfied {
+        required_level: GreenLevel,
+        observed_level: GreenLevel,
+    },
+    Unsatisfied {
+        required_level: GreenLevel,
+        observed_level: Option<GreenLevel>,
+    },
+}
+
+impl GreenContractOutcome {
+    #[must_use]
+    pub fn is_satisfied(&self) -> bool {
+        matches!(self, Self::Satisfied { .. })
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn given_matching_level_when_evaluating_contract_then_it_is_satisfied() {
+        // given
+        let contract = GreenContract::new(GreenLevel::Package);
+
+        // when
+        let outcome = contract.evaluate(Some(GreenLevel::Package));
+
+        // then
+        assert_eq!(
+            outcome,
+            GreenContractOutcome::Satisfied {
+                required_level: GreenLevel::Package,
+                observed_level: GreenLevel::Package,
+            }
+        );
+        assert!(outcome.is_satisfied());
+    }
+
+    #[test]
+    fn given_higher_level_when_checking_requirement_then_it_still_satisfies_contract() {
+        // given
+        let contract = GreenContract::new(GreenLevel::TargetedTests);
+
+        // when
+        let is_satisfied = contract.is_satisfied_by(GreenLevel::Workspace);
+
+        // then
+        assert!(is_satisfied);
+    }
+
+    #[test]
+    fn given_lower_level_when_evaluating_contract_then_it_is_unsatisfied() {
+        // given
+        let contract = GreenContract::new(GreenLevel::Workspace);
+
+        // when
+        let outcome = contract.evaluate(Some(GreenLevel::Package));
+
+        // then
+        assert_eq!(
+            outcome,
+            GreenContractOutcome::Unsatisfied {
+                required_level: GreenLevel::Workspace,
+                observed_level: Some(GreenLevel::Package),
+            }
+        );
+        assert!(!outcome.is_satisfied());
+    }
+
+    #[test]
+    fn given_no_green_level_when_evaluating_contract_then_contract_is_unsatisfied() {
+        // given
+        let contract = GreenContract::new(GreenLevel::MergeReady);
+
+        // when
+        let outcome = contract.evaluate(None);
+
+        // then
+        assert_eq!(
+            outcome,
+            GreenContractOutcome::Unsatisfied {
+                required_level: GreenLevel::MergeReady,
+                observed_level: None,
+            }
+        );
+    }
+}
@@ -0,0 +1,383 @@
+#![allow(clippy::similar_names)]
+use serde::{Deserialize, Serialize};
+use serde_json::Value;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+pub enum LaneEventName {
+    #[serde(rename = "lane.started")]
+    Started,
+    #[serde(rename = "lane.ready")]
+    Ready,
+    #[serde(rename = "lane.prompt_misdelivery")]
+    PromptMisdelivery,
+    #[serde(rename = "lane.blocked")]
+    Blocked,
+    #[serde(rename = "lane.red")]
+    Red,
+    #[serde(rename = "lane.green")]
+    Green,
+    #[serde(rename = "lane.commit.created")]
+    CommitCreated,
+    #[serde(rename = "lane.pr.opened")]
+    PrOpened,
+    #[serde(rename = "lane.merge.ready")]
+    MergeReady,
+    #[serde(rename = "lane.finished")]
+    Finished,
+    #[serde(rename = "lane.failed")]
+    Failed,
+    #[serde(rename = "lane.reconciled")]
+    Reconciled,
+    #[serde(rename = "lane.merged")]
+    Merged,
+    #[serde(rename = "lane.superseded")]
+    Superseded,
+    #[serde(rename = "lane.closed")]
+    Closed,
+    #[serde(rename = "branch.stale_against_main")]
+    BranchStaleAgainstMain,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum LaneEventStatus {
+    Running,
+    Ready,
+    Blocked,
+    Red,
+    Green,
+    Completed,
+    Failed,
+    Reconciled,
+    Merged,
+    Superseded,
+    Closed,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum LaneFailureClass {
+    PromptDelivery,
+    TrustGate,
+    BranchDivergence,
+    Compile,
+    Test,
+    PluginStartup,
+    McpStartup,
+    McpHandshake,
+    GatewayRouting,
+    ToolRuntime,
+    Infra,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct LaneEventBlocker {
+    #[serde(rename = "failureClass")]
+    pub failure_class: LaneFailureClass,
+    pub detail: String,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct LaneCommitProvenance {
+    pub commit: String,
+    pub branch: String,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub worktree: Option<String>,
+    #[serde(rename = "canonicalCommit", skip_serializing_if = "Option::is_none")]
+    pub canonical_commit: Option<String>,
+    #[serde(rename = "supersededBy", skip_serializing_if = "Option::is_none")]
+    pub superseded_by: Option<String>,
+    #[serde(default, skip_serializing_if = "Vec::is_empty")]
+    pub lineage: Vec<String>,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct LaneEvent {
+    pub event: LaneEventName,
+    pub status: LaneEventStatus,
+    #[serde(rename = "emittedAt")]
+    pub emitted_at: String,
+    #[serde(rename = "failureClass", skip_serializing_if = "Option::is_none")]
+    pub failure_class: Option<LaneFailureClass>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub detail: Option<String>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub data: Option<Value>,
+}
+
+impl LaneEvent {
+    #[must_use]
+    pub fn new(
+        event: LaneEventName,
+        status: LaneEventStatus,
+        emitted_at: impl Into<String>,
+    ) -> Self {
+        Self {
+            event,
+            status,
+            emitted_at: emitted_at.into(),
+            failure_class: None,
+            detail: None,
+            data: None,
+        }
+    }
+
+    #[must_use]
+    pub fn started(emitted_at: impl Into<String>) -> Self {
+        Self::new(LaneEventName::Started, LaneEventStatus::Running, emitted_at)
+    }
+
+    #[must_use]
+    pub fn finished(emitted_at: impl Into<String>, detail: Option<String>) -> Self {
+        Self::new(
+            LaneEventName::Finished,
+            LaneEventStatus::Completed,
+            emitted_at,
+        )
+        .with_optional_detail(detail)
+    }
+
+    #[must_use]
+    pub fn commit_created(
+        emitted_at: impl Into<String>,
+        detail: Option<String>,
+        provenance: LaneCommitProvenance,
+    ) -> Self {
+        Self::new(
+            LaneEventName::CommitCreated,
+            LaneEventStatus::Completed,
+            emitted_at,
+        )
+        .with_optional_detail(detail)
+        .with_data(serde_json::to_value(provenance).expect("commit provenance should serialize"))
+    }
+
+    #[must_use]
+    pub fn superseded(
+        emitted_at: impl Into<String>,
+        detail: Option<String>,
+        provenance: LaneCommitProvenance,
+    ) -> Self {
+        Self::new(
+            LaneEventName::Superseded,
+            LaneEventStatus::Superseded,
+            emitted_at,
+        )
+        .with_optional_detail(detail)
+        .with_data(serde_json::to_value(provenance).expect("commit provenance should serialize"))
+    }
+
+    #[must_use]
+    pub fn blocked(emitted_at: impl Into<String>, blocker: &LaneEventBlocker) -> Self {
+        Self::new(LaneEventName::Blocked, LaneEventStatus::Blocked, emitted_at)
+            .with_failure_class(blocker.failure_class)
+            .with_detail(blocker.detail.clone())
+    }
+
+    #[must_use]
+    pub fn failed(emitted_at: impl Into<String>, blocker: &LaneEventBlocker) -> Self {
+        Self::new(LaneEventName::Failed, LaneEventStatus::Failed, emitted_at)
+            .with_failure_class(blocker.failure_class)
+            .with_detail(blocker.detail.clone())
+    }
+
+    #[must_use]
+    pub fn with_failure_class(mut self, failure_class: LaneFailureClass) -> Self {
+        self.failure_class = Some(failure_class);
+        self
+    }
+
+    #[must_use]
+    pub fn with_detail(mut self, detail: impl Into<String>) -> Self {
+        self.detail = Some(detail.into());
+        self
+    }
+
+    #[must_use]
+    pub fn with_optional_detail(mut self, detail: Option<String>) -> Self {
+        self.detail = detail;
+        self
+    }
+
+    #[must_use]
+    pub fn with_data(mut self, data: Value) -> Self {
+        self.data = Some(data);
+        self
+    }
+}
+
+#[must_use]
+pub fn dedupe_superseded_commit_events(events: &[LaneEvent]) -> Vec<LaneEvent> {
+    let mut keep = vec![true; events.len()];
+    let mut latest_by_key = std::collections::BTreeMap::<String, usize>::new();
+
+    for (index, event) in events.iter().enumerate() {
+        if event.event != LaneEventName::CommitCreated {
+            continue;
+        }
+        let Some(data) = event.data.as_ref() else {
+            continue;
+        };
+        let key = data
+            .get("canonicalCommit")
+            .or_else(|| data.get("commit"))
+            .and_then(serde_json::Value::as_str)
+            .map(str::to_string);
+        let superseded = data
+            .get("supersededBy")
+            .and_then(serde_json::Value::as_str)
+            .is_some();
+        if superseded {
+            keep[index] = false;
+            continue;
+        }
+        if let Some(key) = key {
+            if let Some(previous) = latest_by_key.insert(key, index) {
+                keep[previous] = false;
+            }
+        }
+    }
+
+    events
+        .iter()
+        .cloned()
+        .zip(keep)
+        .filter_map(|(event, retain)| retain.then_some(event))
+        .collect()
+}
+
+#[cfg(test)]
+mod tests {
+    use serde_json::json;
+
+    use super::{
+        dedupe_superseded_commit_events, LaneCommitProvenance, LaneEvent, LaneEventBlocker,
+        LaneEventName, LaneEventStatus, LaneFailureClass,
+    };
+
+    #[test]
+    fn canonical_lane_event_names_serialize_to_expected_wire_values() {
+        let cases = [
+            (LaneEventName::Started, "lane.started"),
+            (LaneEventName::Ready, "lane.ready"),
+            (LaneEventName::PromptMisdelivery, "lane.prompt_misdelivery"),
+            (LaneEventName::Blocked, "lane.blocked"),
+            (LaneEventName::Red, "lane.red"),
+            (LaneEventName::Green, "lane.green"),
+            (LaneEventName::CommitCreated, "lane.commit.created"),
+            (LaneEventName::PrOpened, "lane.pr.opened"),
+            (LaneEventName::MergeReady, "lane.merge.ready"),
+            (LaneEventName::Finished, "lane.finished"),
+            (LaneEventName::Failed, "lane.failed"),
+            (LaneEventName::Reconciled, "lane.reconciled"),
+            (LaneEventName::Merged, "lane.merged"),
+            (LaneEventName::Superseded, "lane.superseded"),
+            (LaneEventName::Closed, "lane.closed"),
+            (
+                LaneEventName::BranchStaleAgainstMain,
+                "branch.stale_against_main",
+            ),
+        ];
+
+        for (event, expected) in cases {
+            assert_eq!(
+                serde_json::to_value(event).expect("serialize event"),
+                json!(expected)
+            );
+        }
+    }
+
+    #[test]
+    fn failure_classes_cover_canonical_taxonomy_wire_values() {
+        let cases = [
+            (LaneFailureClass::PromptDelivery, "prompt_delivery"),
+            (LaneFailureClass::TrustGate, "trust_gate"),
+            (LaneFailureClass::BranchDivergence, "branch_divergence"),
+            (LaneFailureClass::Compile, "compile"),
+            (LaneFailureClass::Test, "test"),
+            (LaneFailureClass::PluginStartup, "plugin_startup"),
+            (LaneFailureClass::McpStartup, "mcp_startup"),
+            (LaneFailureClass::McpHandshake, "mcp_handshake"),
+            (LaneFailureClass::GatewayRouting, "gateway_routing"),
+            (LaneFailureClass::ToolRuntime, "tool_runtime"),
+            (LaneFailureClass::Infra, "infra"),
+        ];
+
+        for (failure_class, expected) in cases {
+            assert_eq!(
+                serde_json::to_value(failure_class).expect("serialize failure class"),
+                json!(expected)
+            );
+        }
+    }
+
+    #[test]
+    fn blocked_and_failed_events_reuse_blocker_details() {
+        let blocker = LaneEventBlocker {
+            failure_class: LaneFailureClass::McpStartup,
+            detail: "broken server".to_string(),
+        };
+
+        let blocked = LaneEvent::blocked("2026-04-04T00:00:00Z", &blocker);
+        let failed = LaneEvent::failed("2026-04-04T00:00:01Z", &blocker);
+
+        assert_eq!(blocked.event, LaneEventName::Blocked);
+        assert_eq!(blocked.status, LaneEventStatus::Blocked);
+        assert_eq!(blocked.failure_class, Some(LaneFailureClass::McpStartup));
+        assert_eq!(failed.event, LaneEventName::Failed);
+        assert_eq!(failed.status, LaneEventStatus::Failed);
+        assert_eq!(failed.detail.as_deref(), Some("broken server"));
+    }
+
+    #[test]
+    fn commit_events_can_carry_worktree_and_supersession_metadata() {
+        let event = LaneEvent::commit_created(
+            "2026-04-04T00:00:00Z",
+            Some("commit created".to_string()),
+            LaneCommitProvenance {
+                commit: "abc123".to_string(),
+                branch: "feature/provenance".to_string(),
+                worktree: Some("wt-a".to_string()),
+                canonical_commit: Some("abc123".to_string()),
+                superseded_by: None,
+                lineage: vec!["abc123".to_string()],
+            },
+        );
+        let event_json = serde_json::to_value(&event).expect("lane event should serialize");
+        assert_eq!(event_json["event"], "lane.commit.created");
+        assert_eq!(event_json["data"]["branch"], "feature/provenance");
+        assert_eq!(event_json["data"]["worktree"], "wt-a");
+    }
+
+    #[test]
+    fn dedupes_superseded_commit_events_by_canonical_commit() {
+        let retained = dedupe_superseded_commit_events(&[
+            LaneEvent::commit_created(
+                "2026-04-04T00:00:00Z",
+                Some("old".to_string()),
+                LaneCommitProvenance {
+                    commit: "old123".to_string(),
+                    branch: "feature/provenance".to_string(),
+                    worktree: Some("wt-a".to_string()),
+                    canonical_commit: Some("canon123".to_string()),
+                    superseded_by: Some("new123".to_string()),
+                    lineage: vec!["old123".to_string(), "new123".to_string()],
+                },
+            ),
+            LaneEvent::commit_created(
+                "2026-04-04T00:00:01Z",
+                Some("new".to_string()),
+                LaneCommitProvenance {
+                    commit: "new123".to_string(),
+                    branch: "feature/provenance".to_string(),
+                    worktree: Some("wt-b".to_string()),
+                    canonical_commit: Some("canon123".to_string()),
+                    superseded_by: None,
+                    lineage: vec!["old123".to_string(), "new123".to_string()],
+                },
+            ),
+        ]);
+        assert_eq!(retained.len(), 1);
+        assert_eq!(retained[0].detail.as_deref(), Some("new"));
+    }
+}
@@ -1,64 +1,112 @@
+//! Core runtime primitives for the `claw` CLI and supporting crates.
+//!
+//! This crate owns session persistence, permission evaluation, prompt assembly,
+//! MCP plumbing, tool-facing file operations, and the core conversation loop
+//! that drives interactive and one-shot turns.
+
 mod bash;
+pub mod bash_validation;
 mod bootstrap;
+pub mod branch_lock;
 mod compact;
 mod config;
+pub mod config_validate;
 mod conversation;
 mod file_ops;
+mod git_context;
+pub mod green_contract;
 mod hooks;
 mod json;
+mod lane_events;
+pub mod lsp_client;
 mod mcp;
 mod mcp_client;
+pub mod mcp_lifecycle_hardened;
+pub mod mcp_server;
 mod mcp_stdio;
+pub mod mcp_tool_bridge;
 mod oauth;
+pub mod permission_enforcer;
 mod permissions;
+pub mod plugin_lifecycle;
+mod policy_engine;
 mod prompt;
+pub mod recovery_recipes;
 mod remote;
 pub mod sandbox;
 mod session;
+pub mod session_control;
+pub use session_control::SessionStore;
+mod sse;
+pub mod stale_base;
+pub mod stale_branch;
+pub mod summary_compression;
+pub mod task_packet;
+pub mod task_registry;
+pub mod team_cron_registry;
+#[cfg(test)]
+mod trust_resolver;
 mod usage;
+pub mod worker_boot;

-pub use lsp::{
-    FileDiagnostics, LspContextEnrichment, LspError, LspManager, LspServerConfig,
-    SymbolLocation, WorkspaceDiagnostics,
-};
 pub use bash::{execute_bash, BashCommandInput, BashCommandOutput};
 pub use bootstrap::{BootstrapPhase, BootstrapPlan};
+pub use branch_lock::{detect_branch_lock_collisions, BranchLockCollision, BranchLockIntent};
 pub use compact::{
    compact_session, estimate_session_tokens, format_compact_summary,
    get_compact_continuation_message, should_compact, CompactionConfig, CompactionResult,
 };
 pub use config::{
-    ConfigEntry, ConfigError, ConfigLoader, ConfigSource, McpManagedProxyServerConfig,
-    McpConfigCollection, McpOAuthConfig, McpRemoteServerConfig, McpSdkServerConfig,
+    ConfigEntry, ConfigError, ConfigLoader, ConfigSource, McpConfigCollection,
+    McpManagedProxyServerConfig, McpOAuthConfig, McpRemoteServerConfig, McpSdkServerConfig,
    McpServerConfig, McpStdioServerConfig, McpTransport, McpWebSocketServerConfig, OAuthConfig,
-    ResolvedPermissionMode, RuntimeConfig, RuntimeFeatureConfig, RuntimeHookConfig,
-    RuntimePluginConfig, ScopedMcpServerConfig, CLAW_SETTINGS_SCHEMA_NAME,
+    ProviderFallbackConfig, ResolvedPermissionMode, RuntimeConfig, RuntimeFeatureConfig,
+    RuntimeHookConfig, RuntimePermissionRuleConfig, RuntimePluginConfig, ScopedMcpServerConfig,
+    CLAW_SETTINGS_SCHEMA_NAME,
+};
+pub use config_validate::{
+    check_unsupported_format, format_diagnostics, validate_config_file, ConfigDiagnostic,
+    DiagnosticKind, ValidationResult,
 };
 pub use conversation::{
-    ApiClient, ApiRequest, AssistantEvent, ConversationRuntime, RuntimeError, StaticToolExecutor,
-    ToolError, ToolExecutor, TurnSummary,
+    auto_compaction_threshold_from_env, ApiClient, ApiRequest, AssistantEvent, AutoCompactionEvent,
+    ConversationRuntime, PromptCacheEvent, RuntimeError, StaticToolExecutor, ToolError,
+    ToolExecutor, TurnSummary,
 };
 pub use file_ops::{
    edit_file, glob_search, grep_search, read_file, write_file, EditFileOutput, GlobSearchOutput,
    GrepSearchInput, GrepSearchOutput, ReadFileOutput, StructuredPatchHunk, TextFilePayload,
    WriteFileOutput,
 };
-pub use hooks::{HookEvent, HookRunResult, HookRunner};
+pub use git_context::{GitCommitEntry, GitContext};
+pub use hooks::{
+    HookAbortSignal, HookEvent, HookProgressEvent, HookProgressReporter, HookRunResult, HookRunner,
+};
+pub use lane_events::{
+    dedupe_superseded_commit_events, LaneCommitProvenance, LaneEvent, LaneEventBlocker,
+    LaneEventName, LaneEventStatus, LaneFailureClass,
+};
 pub use mcp::{
    mcp_server_signature, mcp_tool_name, mcp_tool_prefix, normalize_name_for_mcp,
    scoped_mcp_config_hash, unwrap_ccr_proxy_url,
 };
 pub use mcp_client::{
-    McpManagedProxyTransport, McpClientAuth, McpClientBootstrap, McpClientTransport,
+    McpClientAuth, McpClientBootstrap, McpClientTransport, McpManagedProxyTransport,
    McpRemoteTransport, McpSdkTransport, McpStdioTransport,
 };
+pub use mcp_lifecycle_hardened::{
+    McpDegradedReport, McpErrorSurface, McpFailedServer, McpLifecyclePhase, McpLifecycleState,
+    McpLifecycleValidator, McpPhaseResult,
+};
+pub use mcp_server::{McpServer, McpServerSpec, ToolCallHandler, MCP_SERVER_PROTOCOL_VERSION};
 pub use mcp_stdio::{
    spawn_mcp_stdio_process, JsonRpcError, JsonRpcId, JsonRpcRequest, JsonRpcResponse,
-    ManagedMcpTool, McpInitializeClientInfo, McpInitializeParams, McpInitializeResult,
-    McpInitializeServerInfo, McpListResourcesParams, McpListResourcesResult, McpListToolsParams,
-    McpListToolsResult, McpReadResourceParams, McpReadResourceResult, McpResource,
-    McpResourceContents, McpServerManager, McpServerManagerError, McpStdioProcess, McpTool,
-    McpToolCallContent, McpToolCallParams, McpToolCallResult, UnsupportedMcpServer,
+    ManagedMcpTool, McpDiscoveryFailure, McpInitializeClientInfo, McpInitializeParams,
+    McpInitializeResult, McpInitializeServerInfo, McpListResourcesParams, McpListResourcesResult,
+    McpListToolsParams, McpListToolsResult, McpReadResourceParams, McpReadResourceResult,
+    McpResource, McpResourceContents, McpServerManager, McpServerManagerError, McpStdioProcess,
+    McpTool, McpToolCallContent, McpToolCallParams, McpToolCallResult, McpToolDiscoveryReport,
+    UnsupportedMcpServer,
 };
 pub use oauth::{
    clear_oauth_credentials, code_challenge_s256, credentials_path, generate_pkce_pair,
@@ -68,22 +116,59 @@ pub use oauth::{
    PkceChallengeMethod, PkceCodePair,
 };
 pub use permissions::{
-    PermissionMode, PermissionOutcome, PermissionPolicy, PermissionPromptDecision,
-    PermissionPrompter, PermissionRequest,
+    PermissionContext, PermissionMode, PermissionOutcome, PermissionOverride, PermissionPolicy,
+    PermissionPromptDecision, PermissionPrompter, PermissionRequest,
+};
+pub use plugin_lifecycle::{
+    DegradedMode, DiscoveryResult, PluginHealthcheck, PluginLifecycle, PluginLifecycleEvent,
+    PluginState, ResourceInfo, ServerHealth, ServerStatus, ToolInfo,
+};
+pub use policy_engine::{
+    evaluate, DiffScope, GreenLevel, LaneBlocker, LaneContext, PolicyAction, PolicyCondition,
+    PolicyEngine, PolicyRule, ReconcileReason, ReviewStatus,
 };
 pub use prompt::{
    load_system_prompt, prepend_bullets, ContextFile, ProjectContext, PromptBuildError,
    SystemPromptBuilder, FRONTIER_MODEL_NAME, SYSTEM_PROMPT_DYNAMIC_BOUNDARY,
 };
+pub use recovery_recipes::{
+    attempt_recovery, recipe_for, EscalationPolicy, FailureScenario, RecoveryContext,
+    RecoveryEvent, RecoveryRecipe, RecoveryResult, RecoveryStep,
+};
 pub use remote::{
    inherited_upstream_proxy_env, no_proxy_list, read_token, upstream_proxy_ws_url,
    RemoteSessionContext, UpstreamProxyBootstrap, UpstreamProxyState, DEFAULT_REMOTE_BASE_URL,
    DEFAULT_SESSION_TOKEN_PATH, DEFAULT_SYSTEM_CA_BUNDLE, NO_PROXY_HOSTS, UPSTREAM_PROXY_ENV_KEYS,
 };
-pub use session::{ContentBlock, ConversationMessage, MessageRole, Session, SessionError};
+pub use sandbox::{
+    build_linux_sandbox_command, detect_container_environment, detect_container_environment_from,
+    resolve_sandbox_status, resolve_sandbox_status_for_request, ContainerEnvironment,
+    FilesystemIsolationMode, LinuxSandboxCommand, SandboxConfig, SandboxDetectionInputs,
+    SandboxRequest, SandboxStatus,
+};
+pub use session::{
+    ContentBlock, ConversationMessage, MessageRole, Session, SessionCompaction, SessionError,
+    SessionFork, SessionPromptEntry,
+};
+pub use sse::{IncrementalSseParser, SseEvent};
+pub use stale_base::{
+    check_base_commit, format_stale_base_warning, read_claw_base_file, resolve_expected_base,
+    BaseCommitSource, BaseCommitState,
+};
+pub use stale_branch::{
+    apply_policy, check_freshness, BranchFreshness, StaleBranchAction, StaleBranchEvent,
+    StaleBranchPolicy,
+};
+pub use task_packet::{validate_packet, TaskPacket, TaskPacketValidationError, ValidatedPacket};
+#[cfg(test)]
+pub use trust_resolver::{TrustConfig, TrustDecision, TrustEvent, TrustPolicy, TrustResolver};
 pub use usage::{
    format_usd, pricing_for_model, ModelPricing, TokenUsage, UsageCostEstimate, UsageTracker,
 };
+pub use worker_boot::{
+    Worker, WorkerEvent, WorkerEventKind, WorkerEventPayload, WorkerFailure, WorkerFailureKind,
+    WorkerPromptTarget, WorkerReadySnapshot, WorkerRegistry, WorkerStatus, WorkerTrustResolution,
+};

 #[cfg(test)]
 pub(crate) fn test_env_lock() -> std::sync::MutexGuard<'static, ()> {
@@ -0,0 +1,747 @@
+#![allow(clippy::should_implement_trait, clippy::must_use_candidate)]
+//! LSP (Language Server Protocol) client registry for tool dispatch.
+
+use std::collections::HashMap;
+use std::sync::{Arc, Mutex};
+
+use serde::{Deserialize, Serialize};
+
+/// Supported LSP actions.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum LspAction {
+    Diagnostics,
+    Hover,
+    Definition,
+    References,
+    Completion,
+    Symbols,
+    Format,
+}
+
+impl LspAction {
+    pub fn from_str(s: &str) -> Option<Self> {
+        match s {
+            "diagnostics" => Some(Self::Diagnostics),
+            "hover" => Some(Self::Hover),
+            "definition" | "goto_definition" => Some(Self::Definition),
+            "references" | "find_references" => Some(Self::References),
+            "completion" | "completions" => Some(Self::Completion),
+            "symbols" | "document_symbols" => Some(Self::Symbols),
+            "format" | "formatting" => Some(Self::Format),
+            _ => None,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct LspDiagnostic {
+    pub path: String,
+    pub line: u32,
+    pub character: u32,
+    pub severity: String,
+    pub message: String,
+    pub source: Option<String>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct LspLocation {
+    pub path: String,
+    pub line: u32,
+    pub character: u32,
+    pub end_line: Option<u32>,
+    pub end_character: Option<u32>,
+    pub preview: Option<String>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct LspHoverResult {
+    pub content: String,
+    pub language: Option<String>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct LspCompletionItem {
+    pub label: String,
+    pub kind: Option<String>,
+    pub detail: Option<String>,
+    pub insert_text: Option<String>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct LspSymbol {
+    pub name: String,
+    pub kind: String,
+    pub path: String,
+    pub line: u32,
+    pub character: u32,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum LspServerStatus {
+    Connected,
+    Disconnected,
+    Starting,
+    Error,
+}
+
+impl std::fmt::Display for LspServerStatus {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Connected => write!(f, "connected"),
+            Self::Disconnected => write!(f, "disconnected"),
+            Self::Starting => write!(f, "starting"),
+            Self::Error => write!(f, "error"),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct LspServerState {
+    pub language: String,
+    pub status: LspServerStatus,
+    pub root_path: Option<String>,
+    pub capabilities: Vec<String>,
+    pub diagnostics: Vec<LspDiagnostic>,
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct LspRegistry {
+    inner: Arc<Mutex<RegistryInner>>,
+}
+
+#[derive(Debug, Default)]
+struct RegistryInner {
+    servers: HashMap<String, LspServerState>,
+}
+
+impl LspRegistry {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    pub fn register(
+        &self,
+        language: &str,
+        status: LspServerStatus,
+        root_path: Option<&str>,
+        capabilities: Vec<String>,
+    ) {
+        let mut inner = self.inner.lock().expect("lsp registry lock poisoned");
+        inner.servers.insert(
+            language.to_owned(),
+            LspServerState {
+                language: language.to_owned(),
+                status,
+                root_path: root_path.map(str::to_owned),
+                capabilities,
+                diagnostics: Vec::new(),
+            },
+        );
+    }
+
+    pub fn get(&self, language: &str) -> Option<LspServerState> {
+        let inner = self.inner.lock().expect("lsp registry lock poisoned");
+        inner.servers.get(language).cloned()
+    }
+
+    /// Find the appropriate server for a file path based on extension.
+    pub fn find_server_for_path(&self, path: &str) -> Option<LspServerState> {
+        let ext = std::path::Path::new(path)
+            .extension()
+            .and_then(|e| e.to_str())
+            .unwrap_or("");
+
+        let language = match ext {
+            "rs" => "rust",
+            "ts" | "tsx" => "typescript",
+            "js" | "jsx" => "javascript",
+            "py" => "python",
+            "go" => "go",
+            "java" => "java",
+            "c" | "h" => "c",
+            "cpp" | "hpp" | "cc" => "cpp",
+            "rb" => "ruby",
+            "lua" => "lua",
+            _ => return None,
+        };
+
+        self.get(language)
+    }
+
+    /// List all registered servers.
+    pub fn list_servers(&self) -> Vec<LspServerState> {
+        let inner = self.inner.lock().expect("lsp registry lock poisoned");
+        inner.servers.values().cloned().collect()
+    }
+
+    /// Add diagnostics to a server.
+    pub fn add_diagnostics(
+        &self,
+        language: &str,
+        diagnostics: Vec<LspDiagnostic>,
+    ) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("lsp registry lock poisoned");
+        let server = inner
+            .servers
+            .get_mut(language)
+            .ok_or_else(|| format!("LSP server not found for language: {language}"))?;
+        server.diagnostics.extend(diagnostics);
+        Ok(())
+    }
+
+    /// Get diagnostics for a specific file path.
+    pub fn get_diagnostics(&self, path: &str) -> Vec<LspDiagnostic> {
+        let inner = self.inner.lock().expect("lsp registry lock poisoned");
+        inner
+            .servers
+            .values()
+            .flat_map(|s| &s.diagnostics)
+            .filter(|d| d.path == path)
+            .cloned()
+            .collect()
+    }
+
+    /// Clear diagnostics for a language server.
+    pub fn clear_diagnostics(&self, language: &str) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("lsp registry lock poisoned");
+        let server = inner
+            .servers
+            .get_mut(language)
+            .ok_or_else(|| format!("LSP server not found for language: {language}"))?;
+        server.diagnostics.clear();
+        Ok(())
+    }
+
+    /// Disconnect a server.
+    pub fn disconnect(&self, language: &str) -> Option<LspServerState> {
+        let mut inner = self.inner.lock().expect("lsp registry lock poisoned");
+        inner.servers.remove(language)
+    }
+
+    #[must_use]
+    pub fn len(&self) -> usize {
+        let inner = self.inner.lock().expect("lsp registry lock poisoned");
+        inner.servers.len()
+    }
+
+    #[must_use]
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+
+    /// Dispatch an LSP action and return a structured result.
+    pub fn dispatch(
+        &self,
+        action: &str,
+        path: Option<&str>,
+        line: Option<u32>,
+        character: Option<u32>,
+        _query: Option<&str>,
+    ) -> Result<serde_json::Value, String> {
+        let lsp_action =
+            LspAction::from_str(action).ok_or_else(|| format!("unknown LSP action: {action}"))?;
+
+        // For diagnostics, we can check existing cached diagnostics
+        if lsp_action == LspAction::Diagnostics {
+            if let Some(path) = path {
+                let diags = self.get_diagnostics(path);
+                return Ok(serde_json::json!({
+                    "action": "diagnostics",
+                    "path": path,
+                    "diagnostics": diags,
+                    "count": diags.len()
+                }));
+            }
+            // All diagnostics across all servers
+            let inner = self.inner.lock().expect("lsp registry lock poisoned");
+            let all_diags: Vec<_> = inner
+                .servers
+                .values()
+                .flat_map(|s| &s.diagnostics)
+                .collect();
+            return Ok(serde_json::json!({
+                "action": "diagnostics",
+                "diagnostics": all_diags,
+                "count": all_diags.len()
+            }));
+        }
+
+        // For other actions, we need a connected server for the given file
+        let path = path.ok_or("path is required for this LSP action")?;
+        let server = self
+            .find_server_for_path(path)
+            .ok_or_else(|| format!("no LSP server available for path: {path}"))?;
+
+        if server.status != LspServerStatus::Connected {
+            return Err(format!(
+                "LSP server for '{}' is not connected (status: {})",
+                server.language, server.status
+            ));
+        }
+
+        // Return structured placeholder — actual LSP JSON-RPC calls would
+        // go through the real LSP process here.
+        Ok(serde_json::json!({
+            "action": action,
+            "path": path,
+            "line": line,
+            "character": character,
+            "language": server.language,
+            "status": "dispatched",
+            "message": format!("LSP {} dispatched to {} server", action, server.language)
+        }))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn registers_and_retrieves_server() {
+        let registry = LspRegistry::new();
+        registry.register(
+            "rust",
+            LspServerStatus::Connected,
+            Some("/workspace"),
+            vec!["hover".into(), "completion".into()],
+        );
+
+        let server = registry.get("rust").expect("should exist");
+        assert_eq!(server.language, "rust");
+        assert_eq!(server.status, LspServerStatus::Connected);
+        assert_eq!(server.capabilities.len(), 2);
+    }
+
+    #[test]
+    fn finds_server_by_file_extension() {
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+        registry.register("typescript", LspServerStatus::Connected, None, vec![]);
+
+        let rs_server = registry.find_server_for_path("src/main.rs").unwrap();
+        assert_eq!(rs_server.language, "rust");
+
+        let ts_server = registry.find_server_for_path("src/index.ts").unwrap();
+        assert_eq!(ts_server.language, "typescript");
+
+        assert!(registry.find_server_for_path("data.csv").is_none());
+    }
+
+    #[test]
+    fn manages_diagnostics() {
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+
+        registry
+            .add_diagnostics(
+                "rust",
+                vec![LspDiagnostic {
+                    path: "src/main.rs".into(),
+                    line: 10,
+                    character: 5,
+                    severity: "error".into(),
+                    message: "mismatched types".into(),
+                    source: Some("rust-analyzer".into()),
+                }],
+            )
+            .unwrap();
+
+        let diags = registry.get_diagnostics("src/main.rs");
+        assert_eq!(diags.len(), 1);
+        assert_eq!(diags[0].message, "mismatched types");
+
+        registry.clear_diagnostics("rust").unwrap();
+        assert!(registry.get_diagnostics("src/main.rs").is_empty());
+    }
+
+    #[test]
+    fn dispatches_diagnostics_action() {
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+        registry
+            .add_diagnostics(
+                "rust",
+                vec![LspDiagnostic {
+                    path: "src/lib.rs".into(),
+                    line: 1,
+                    character: 0,
+                    severity: "warning".into(),
+                    message: "unused import".into(),
+                    source: None,
+                }],
+            )
+            .unwrap();
+
+        let result = registry
+            .dispatch("diagnostics", Some("src/lib.rs"), None, None, None)
+            .unwrap();
+        assert_eq!(result["count"], 1);
+    }
+
+    #[test]
+    fn dispatches_hover_action() {
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+
+        let result = registry
+            .dispatch("hover", Some("src/main.rs"), Some(10), Some(5), None)
+            .unwrap();
+        assert_eq!(result["action"], "hover");
+        assert_eq!(result["language"], "rust");
+    }
+
+    #[test]
+    fn rejects_action_on_disconnected_server() {
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Disconnected, None, vec![]);
+
+        assert!(registry
+            .dispatch("hover", Some("src/main.rs"), Some(1), Some(0), None)
+            .is_err());
+    }
+
+    #[test]
+    fn rejects_unknown_action() {
+        let registry = LspRegistry::new();
+        assert!(registry
+            .dispatch("unknown_action", Some("file.rs"), None, None, None)
+            .is_err());
+    }
+
+    #[test]
+    fn disconnects_server() {
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+        assert_eq!(registry.len(), 1);
+
+        let removed = registry.disconnect("rust");
+        assert!(removed.is_some());
+        assert!(registry.is_empty());
+    }
+
+    #[test]
+    fn lsp_action_from_str_all_aliases() {
+        // given
+        let cases = [
+            ("diagnostics", Some(LspAction::Diagnostics)),
+            ("hover", Some(LspAction::Hover)),
+            ("definition", Some(LspAction::Definition)),
+            ("goto_definition", Some(LspAction::Definition)),
+            ("references", Some(LspAction::References)),
+            ("find_references", Some(LspAction::References)),
+            ("completion", Some(LspAction::Completion)),
+            ("completions", Some(LspAction::Completion)),
+            ("symbols", Some(LspAction::Symbols)),
+            ("document_symbols", Some(LspAction::Symbols)),
+            ("format", Some(LspAction::Format)),
+            ("formatting", Some(LspAction::Format)),
+            ("unknown", None),
+        ];
+
+        // when
+        let resolved: Vec<_> = cases
+            .into_iter()
+            .map(|(input, expected)| (input, LspAction::from_str(input), expected))
+            .collect();
+
+        // then
+        for (input, actual, expected) in resolved {
+            assert_eq!(actual, expected, "unexpected action resolution for {input}");
+        }
+    }
+
+    #[test]
+    fn lsp_server_status_display_all_variants() {
+        // given
+        let cases = [
+            (LspServerStatus::Connected, "connected"),
+            (LspServerStatus::Disconnected, "disconnected"),
+            (LspServerStatus::Starting, "starting"),
+            (LspServerStatus::Error, "error"),
+        ];
+
+        // when
+        let rendered: Vec<_> = cases
+            .into_iter()
+            .map(|(status, expected)| (status.to_string(), expected))
+            .collect();
+
+        // then
+        assert_eq!(
+            rendered,
+            vec![
+                ("connected".to_string(), "connected"),
+                ("disconnected".to_string(), "disconnected"),
+                ("starting".to_string(), "starting"),
+                ("error".to_string(), "error"),
+            ]
+        );
+    }
+
+    #[test]
+    fn dispatch_diagnostics_without_path_aggregates() {
+        // given
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+        registry.register("python", LspServerStatus::Connected, None, vec![]);
+        registry
+            .add_diagnostics(
+                "rust",
+                vec![LspDiagnostic {
+                    path: "src/lib.rs".into(),
+                    line: 1,
+                    character: 0,
+                    severity: "warning".into(),
+                    message: "unused import".into(),
+                    source: Some("rust-analyzer".into()),
+                }],
+            )
+            .expect("rust diagnostics should add");
+        registry
+            .add_diagnostics(
+                "python",
+                vec![LspDiagnostic {
+                    path: "script.py".into(),
+                    line: 2,
+                    character: 4,
+                    severity: "error".into(),
+                    message: "undefined name".into(),
+                    source: Some("pyright".into()),
+                }],
+            )
+            .expect("python diagnostics should add");
+
+        // when
+        let result = registry
+            .dispatch("diagnostics", None, None, None, None)
+            .expect("aggregate diagnostics should work");
+
+        // then
+        assert_eq!(result["action"], "diagnostics");
+        assert_eq!(result["count"], 2);
+        assert_eq!(result["diagnostics"].as_array().map(Vec::len), Some(2));
+    }
+
+    #[test]
+    fn dispatch_non_diagnostics_requires_path() {
+        // given
+        let registry = LspRegistry::new();
+
+        // when
+        let result = registry.dispatch("hover", None, Some(1), Some(0), None);
+
+        // then
+        assert_eq!(
+            result.expect_err("path should be required"),
+            "path is required for this LSP action"
+        );
+    }
+
+    #[test]
+    fn dispatch_no_server_for_path_errors() {
+        // given
+        let registry = LspRegistry::new();
+
+        // when
+        let result = registry.dispatch("hover", Some("notes.md"), Some(1), Some(0), None);
+
+        // then
+        let error = result.expect_err("missing server should fail");
+        assert!(error.contains("no LSP server available for path: notes.md"));
+    }
+
+    #[test]
+    fn dispatch_disconnected_server_error_payload() {
+        // given
+        let registry = LspRegistry::new();
+        registry.register("typescript", LspServerStatus::Disconnected, None, vec![]);
+
+        // when
+        let result = registry.dispatch("hover", Some("src/index.ts"), Some(3), Some(2), None);
+
+        // then
+        let error = result.expect_err("disconnected server should fail");
+        assert!(error.contains("typescript"));
+        assert!(error.contains("disconnected"));
+    }
+
+    #[test]
+    fn find_server_for_all_extensions() {
+        // given
+        let registry = LspRegistry::new();
+        for language in [
+            "rust",
+            "typescript",
+            "javascript",
+            "python",
+            "go",
+            "java",
+            "c",
+            "cpp",
+            "ruby",
+            "lua",
+        ] {
+            registry.register(language, LspServerStatus::Connected, None, vec![]);
+        }
+        let cases = [
+            ("src/main.rs", "rust"),
+            ("src/index.ts", "typescript"),
+            ("src/view.tsx", "typescript"),
+            ("src/app.js", "javascript"),
+            ("src/app.jsx", "javascript"),
+            ("script.py", "python"),
+            ("main.go", "go"),
+            ("Main.java", "java"),
+            ("native.c", "c"),
+            ("native.h", "c"),
+            ("native.cpp", "cpp"),
+            ("native.hpp", "cpp"),
+            ("native.cc", "cpp"),
+            ("script.rb", "ruby"),
+            ("script.lua", "lua"),
+        ];
+
+        // when
+        let resolved: Vec<_> = cases
+            .into_iter()
+            .map(|(path, expected)| {
+                (
+                    path,
+                    registry
+                        .find_server_for_path(path)
+                        .map(|server| server.language),
+                    expected,
+                )
+            })
+            .collect();
+
+        // then
+        for (path, actual, expected) in resolved {
+            assert_eq!(
+                actual.as_deref(),
+                Some(expected),
+                "unexpected mapping for {path}"
+            );
+        }
+    }
+
+    #[test]
+    fn find_server_for_path_no_extension() {
+        // given
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+
+        // when
+        let result = registry.find_server_for_path("Makefile");
+
+        // then
+        assert!(result.is_none());
+    }
+
+    #[test]
+    fn list_servers_with_multiple() {
+        // given
+        let registry = LspRegistry::new();
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+        registry.register("typescript", LspServerStatus::Starting, None, vec![]);
+        registry.register("python", LspServerStatus::Error, None, vec![]);
+
+        // when
+        let servers = registry.list_servers();
+
+        // then
+        assert_eq!(servers.len(), 3);
+        assert!(servers.iter().any(|server| server.language == "rust"));
+        assert!(servers.iter().any(|server| server.language == "typescript"));
+        assert!(servers.iter().any(|server| server.language == "python"));
+    }
+
+    #[test]
+    fn get_missing_server_returns_none() {
+        // given
+        let registry = LspRegistry::new();
+
+        // when
+        let server = registry.get("missing");
+
+        // then
+        assert!(server.is_none());
+    }
+
+    #[test]
+    fn add_diagnostics_missing_language_errors() {
+        // given
+        let registry = LspRegistry::new();
+
+        // when
+        let result = registry.add_diagnostics("missing", vec![]);
+
+        // then
+        let error = result.expect_err("missing language should fail");
+        assert!(error.contains("LSP server not found for language: missing"));
+    }
+
+    #[test]
+    fn get_diagnostics_across_servers() {
+        // given
+        let registry = LspRegistry::new();
+        let shared_path = "shared/file.txt";
+        registry.register("rust", LspServerStatus::Connected, None, vec![]);
+        registry.register("python", LspServerStatus::Connected, None, vec![]);
+        registry
+            .add_diagnostics(
+                "rust",
+                vec![LspDiagnostic {
+                    path: shared_path.into(),
+                    line: 4,
+                    character: 1,
+                    severity: "warning".into(),
+                    message: "warn".into(),
+                    source: None,
+                }],
+            )
+            .expect("rust diagnostics should add");
+        registry
+            .add_diagnostics(
+                "python",
+                vec![LspDiagnostic {
+                    path: shared_path.into(),
+                    line: 8,
+                    character: 3,
+                    severity: "error".into(),
+                    message: "err".into(),
+                    source: None,
+                }],
+            )
+            .expect("python diagnostics should add");
+
+        // when
+        let diagnostics = registry.get_diagnostics(shared_path);
+
+        // then
+        assert_eq!(diagnostics.len(), 2);
+        assert!(diagnostics
+            .iter()
+            .any(|diagnostic| diagnostic.message == "warn"));
+        assert!(diagnostics
+            .iter()
+            .any(|diagnostic| diagnostic.message == "err"));
+    }
+
+    #[test]
+    fn clear_diagnostics_missing_language_errors() {
+        // given
+        let registry = LspRegistry::new();
+
+        // when
+        let result = registry.clear_diagnostics("missing");
+
+        // then
+        let error = result.expect_err("missing language should fail");
+        assert!(error.contains("LSP server not found for language: missing"));
+    }
+}
@@ -84,10 +84,13 @@ pub fn mcp_server_signature(config: &McpServerConfig) -> Option<String> {
 pub fn scoped_mcp_config_hash(config: &ScopedMcpServerConfig) -> String {
    let rendered = match &config.config {
        McpServerConfig::Stdio(stdio) => format!(
-            "stdio|{}|{}|{}",
+            "stdio|{}|{}|{}|{}",
            stdio.command,
            render_command_signature(&stdio.args),
-            render_env_signature(&stdio.env)
+            render_env_signature(&stdio.env),
+            stdio
+                .tool_call_timeout_ms
+                .map_or_else(String::new, |timeout_ms| timeout_ms.to_string())
        ),
        McpServerConfig::Sse(remote) => format!(
            "sse|{}|{}|{}|{}",
@@ -245,6 +248,7 @@ mod tests {
            command: "uvx".to_string(),
            args: vec!["mcp-server".to_string()],
            env: BTreeMap::from([("TOKEN".to_string(), "secret".to_string())]),
+            tool_call_timeout_ms: None,
        });
        assert_eq!(
            mcp_server_signature(&stdio),
@@ -3,6 +3,8 @@ use std::collections::BTreeMap;
 use crate::config::{McpOAuthConfig, McpServerConfig, ScopedMcpServerConfig};
 use crate::mcp::{mcp_server_signature, mcp_tool_prefix, normalize_name_for_mcp};

+pub const DEFAULT_MCP_TOOL_CALL_TIMEOUT_MS: u64 = 60_000;
+
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub enum McpClientTransport {
    Stdio(McpStdioTransport),
@@ -18,6 +20,7 @@ pub struct McpStdioTransport {
    pub command: String,
    pub args: Vec<String>,
    pub env: BTreeMap<String, String>,
+    pub tool_call_timeout_ms: Option<u64>,
 }

 #[derive(Debug, Clone, PartialEq, Eq)]
@@ -75,6 +78,7 @@ impl McpClientTransport {
                command: config.command.clone(),
                args: config.args.clone(),
                env: config.env.clone(),
+                tool_call_timeout_ms: config.tool_call_timeout_ms,
            }),
            McpServerConfig::Sse(config) => Self::Sse(McpRemoteTransport {
                url: config.url.clone(),
@@ -105,6 +109,14 @@ impl McpClientTransport {
    }
 }

+impl McpStdioTransport {
+    #[must_use]
+    pub fn resolved_tool_call_timeout_ms(&self) -> u64 {
+        self.tool_call_timeout_ms
+            .unwrap_or(DEFAULT_MCP_TOOL_CALL_TIMEOUT_MS)
+    }
+}
+
 impl McpClientAuth {
    #[must_use]
    pub fn from_oauth(oauth: Option<McpOAuthConfig>) -> Self {
@@ -136,6 +148,7 @@ mod tests {
                command: "uvx".to_string(),
                args: vec!["mcp-server".to_string()],
                env: BTreeMap::from([("TOKEN".to_string(), "secret".to_string())]),
+                tool_call_timeout_ms: Some(15_000),
            }),
        };

@@ -154,6 +167,7 @@ mod tests {
                    transport.env.get("TOKEN").map(String::as_str),
                    Some("secret")
                );
+                assert_eq!(transport.tool_call_timeout_ms, Some(15_000));
            }
            other => panic!("expected stdio transport, got {other:?}"),
        }
@@ -0,0 +1,843 @@
+#![allow(clippy::unnested_or_patterns, clippy::map_unwrap_or)]
+use std::collections::{BTreeMap, BTreeSet};
+use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
+
+use serde::{Deserialize, Serialize};
+
+fn now_secs() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .unwrap_or_default()
+        .as_secs()
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum McpLifecyclePhase {
+    ConfigLoad,
+    ServerRegistration,
+    SpawnConnect,
+    InitializeHandshake,
+    ToolDiscovery,
+    ResourceDiscovery,
+    Ready,
+    Invocation,
+    ErrorSurfacing,
+    Shutdown,
+    Cleanup,
+}
+
+impl McpLifecyclePhase {
+    #[must_use]
+    pub fn all() -> [Self; 11] {
+        [
+            Self::ConfigLoad,
+            Self::ServerRegistration,
+            Self::SpawnConnect,
+            Self::InitializeHandshake,
+            Self::ToolDiscovery,
+            Self::ResourceDiscovery,
+            Self::Ready,
+            Self::Invocation,
+            Self::ErrorSurfacing,
+            Self::Shutdown,
+            Self::Cleanup,
+        ]
+    }
+}
+
+impl std::fmt::Display for McpLifecyclePhase {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::ConfigLoad => write!(f, "config_load"),
+            Self::ServerRegistration => write!(f, "server_registration"),
+            Self::SpawnConnect => write!(f, "spawn_connect"),
+            Self::InitializeHandshake => write!(f, "initialize_handshake"),
+            Self::ToolDiscovery => write!(f, "tool_discovery"),
+            Self::ResourceDiscovery => write!(f, "resource_discovery"),
+            Self::Ready => write!(f, "ready"),
+            Self::Invocation => write!(f, "invocation"),
+            Self::ErrorSurfacing => write!(f, "error_surfacing"),
+            Self::Shutdown => write!(f, "shutdown"),
+            Self::Cleanup => write!(f, "cleanup"),
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct McpErrorSurface {
+    pub phase: McpLifecyclePhase,
+    pub server_name: Option<String>,
+    pub message: String,
+    pub context: BTreeMap<String, String>,
+    pub recoverable: bool,
+    pub timestamp: u64,
+}
+
+impl McpErrorSurface {
+    #[must_use]
+    pub fn new(
+        phase: McpLifecyclePhase,
+        server_name: Option<String>,
+        message: impl Into<String>,
+        context: BTreeMap<String, String>,
+        recoverable: bool,
+    ) -> Self {
+        Self {
+            phase,
+            server_name,
+            message: message.into(),
+            context,
+            recoverable,
+            timestamp: now_secs(),
+        }
+    }
+}
+
+impl std::fmt::Display for McpErrorSurface {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(
+            f,
+            "MCP lifecycle error during {}: {}",
+            self.phase, self.message
+        )?;
+        if let Some(server_name) = &self.server_name {
+            write!(f, " (server: {server_name})")?;
+        }
+        if !self.context.is_empty() {
+            write!(f, " with context {:?}", self.context)?;
+        }
+        if self.recoverable {
+            write!(f, " [recoverable]")?;
+        }
+        Ok(())
+    }
+}
+
+impl std::error::Error for McpErrorSurface {}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum McpPhaseResult {
+    Success {
+        phase: McpLifecyclePhase,
+        duration: Duration,
+    },
+    Failure {
+        phase: McpLifecyclePhase,
+        error: McpErrorSurface,
+    },
+    Timeout {
+        phase: McpLifecyclePhase,
+        waited: Duration,
+        error: McpErrorSurface,
+    },
+}
+
+impl McpPhaseResult {
+    #[must_use]
+    pub fn phase(&self) -> McpLifecyclePhase {
+        match self {
+            Self::Success { phase, .. }
+            | Self::Failure { phase, .. }
+            | Self::Timeout { phase, .. } => *phase,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct McpLifecycleState {
+    current_phase: Option<McpLifecyclePhase>,
+    phase_errors: BTreeMap<McpLifecyclePhase, Vec<McpErrorSurface>>,
+    phase_timestamps: BTreeMap<McpLifecyclePhase, u64>,
+    phase_results: Vec<McpPhaseResult>,
+}
+
+impl McpLifecycleState {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    #[must_use]
+    pub fn current_phase(&self) -> Option<McpLifecyclePhase> {
+        self.current_phase
+    }
+
+    #[must_use]
+    pub fn errors_for_phase(&self, phase: McpLifecyclePhase) -> &[McpErrorSurface] {
+        self.phase_errors
+            .get(&phase)
+            .map(Vec::as_slice)
+            .unwrap_or(&[])
+    }
+
+    #[must_use]
+    pub fn results(&self) -> &[McpPhaseResult] {
+        &self.phase_results
+    }
+
+    #[must_use]
+    pub fn phase_timestamps(&self) -> &BTreeMap<McpLifecyclePhase, u64> {
+        &self.phase_timestamps
+    }
+
+    #[must_use]
+    pub fn phase_timestamp(&self, phase: McpLifecyclePhase) -> Option<u64> {
+        self.phase_timestamps.get(&phase).copied()
+    }
+
+    fn record_phase(&mut self, phase: McpLifecyclePhase) {
+        self.current_phase = Some(phase);
+        self.phase_timestamps.insert(phase, now_secs());
+    }
+
+    fn record_error(&mut self, error: McpErrorSurface) {
+        self.phase_errors
+            .entry(error.phase)
+            .or_default()
+            .push(error);
+    }
+
+    fn record_result(&mut self, result: McpPhaseResult) {
+        self.phase_results.push(result);
+    }
+
+    fn can_resume_after_error(&self) -> bool {
+        match self.phase_results.last() {
+            Some(McpPhaseResult::Failure { error, .. } | McpPhaseResult::Timeout { error, .. }) => {
+                error.recoverable
+            }
+            _ => false,
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct McpFailedServer {
+    pub server_name: String,
+    pub phase: McpLifecyclePhase,
+    pub error: McpErrorSurface,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct McpDegradedReport {
+    pub working_servers: Vec<String>,
+    pub failed_servers: Vec<McpFailedServer>,
+    pub available_tools: Vec<String>,
+    pub missing_tools: Vec<String>,
+}
+
+impl McpDegradedReport {
+    #[must_use]
+    pub fn new(
+        working_servers: Vec<String>,
+        failed_servers: Vec<McpFailedServer>,
+        available_tools: Vec<String>,
+        expected_tools: Vec<String>,
+    ) -> Self {
+        let working_servers = dedupe_sorted(working_servers);
+        let available_tools = dedupe_sorted(available_tools);
+        let available_tool_set: BTreeSet<_> = available_tools.iter().cloned().collect();
+        let expected_tools = dedupe_sorted(expected_tools);
+        let missing_tools = expected_tools
+            .into_iter()
+            .filter(|tool| !available_tool_set.contains(tool))
+            .collect();
+
+        Self {
+            working_servers,
+            failed_servers,
+            available_tools,
+            missing_tools,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct McpLifecycleValidator {
+    state: McpLifecycleState,
+}
+
+impl McpLifecycleValidator {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    #[must_use]
+    pub fn state(&self) -> &McpLifecycleState {
+        &self.state
+    }
+
+    #[must_use]
+    pub fn validate_phase_transition(from: McpLifecyclePhase, to: McpLifecyclePhase) -> bool {
+        match (from, to) {
+            (McpLifecyclePhase::ConfigLoad, McpLifecyclePhase::ServerRegistration)
+            | (McpLifecyclePhase::ServerRegistration, McpLifecyclePhase::SpawnConnect)
+            | (McpLifecyclePhase::SpawnConnect, McpLifecyclePhase::InitializeHandshake)
+            | (McpLifecyclePhase::InitializeHandshake, McpLifecyclePhase::ToolDiscovery)
+            | (McpLifecyclePhase::ToolDiscovery, McpLifecyclePhase::ResourceDiscovery)
+            | (McpLifecyclePhase::ToolDiscovery, McpLifecyclePhase::Ready)
+            | (McpLifecyclePhase::ResourceDiscovery, McpLifecyclePhase::Ready)
+            | (McpLifecyclePhase::Ready, McpLifecyclePhase::Invocation)
+            | (McpLifecyclePhase::Invocation, McpLifecyclePhase::Ready)
+            | (McpLifecyclePhase::ErrorSurfacing, McpLifecyclePhase::Ready)
+            | (McpLifecyclePhase::ErrorSurfacing, McpLifecyclePhase::Shutdown)
+            | (McpLifecyclePhase::Shutdown, McpLifecyclePhase::Cleanup) => true,
+            (_, McpLifecyclePhase::Shutdown) => from != McpLifecyclePhase::Cleanup,
+            (_, McpLifecyclePhase::ErrorSurfacing) => {
+                from != McpLifecyclePhase::Cleanup && from != McpLifecyclePhase::Shutdown
+            }
+            _ => false,
+        }
+    }
+
+    pub fn run_phase(&mut self, phase: McpLifecyclePhase) -> McpPhaseResult {
+        let started = Instant::now();
+
+        if let Some(current_phase) = self.state.current_phase() {
+            if current_phase == McpLifecyclePhase::ErrorSurfacing
+                && phase == McpLifecyclePhase::Ready
+                && !self.state.can_resume_after_error()
+            {
+                return self.record_failure(McpErrorSurface::new(
+                    phase,
+                    None,
+                    "cannot return to ready after a non-recoverable MCP lifecycle failure",
+                    BTreeMap::from([
+                        ("from".to_string(), current_phase.to_string()),
+                        ("to".to_string(), phase.to_string()),
+                    ]),
+                    false,
+                ));
+            }
+
+            if !Self::validate_phase_transition(current_phase, phase) {
+                return self.record_failure(McpErrorSurface::new(
+                    phase,
+                    None,
+                    format!("invalid MCP lifecycle transition from {current_phase} to {phase}"),
+                    BTreeMap::from([
+                        ("from".to_string(), current_phase.to_string()),
+                        ("to".to_string(), phase.to_string()),
+                    ]),
+                    false,
+                ));
+            }
+        } else if phase != McpLifecyclePhase::ConfigLoad {
+            return self.record_failure(McpErrorSurface::new(
+                phase,
+                None,
+                format!("invalid initial MCP lifecycle phase {phase}"),
+                BTreeMap::from([("phase".to_string(), phase.to_string())]),
+                false,
+            ));
+        }
+
+        self.state.record_phase(phase);
+        let result = McpPhaseResult::Success {
+            phase,
+            duration: started.elapsed(),
+        };
+        self.state.record_result(result.clone());
+        result
+    }
+
+    pub fn record_failure(&mut self, error: McpErrorSurface) -> McpPhaseResult {
+        let phase = error.phase;
+        self.state.record_error(error.clone());
+        self.state.record_phase(McpLifecyclePhase::ErrorSurfacing);
+        let result = McpPhaseResult::Failure { phase, error };
+        self.state.record_result(result.clone());
+        result
+    }
+
+    pub fn record_timeout(
+        &mut self,
+        phase: McpLifecyclePhase,
+        waited: Duration,
+        server_name: Option<String>,
+        mut context: BTreeMap<String, String>,
+    ) -> McpPhaseResult {
+        context.insert("waited_ms".to_string(), waited.as_millis().to_string());
+        let error = McpErrorSurface::new(
+            phase,
+            server_name,
+            format!(
+                "MCP lifecycle phase {phase} timed out after {} ms",
+                waited.as_millis()
+            ),
+            context,
+            true,
+        );
+        self.state.record_error(error.clone());
+        self.state.record_phase(McpLifecyclePhase::ErrorSurfacing);
+        let result = McpPhaseResult::Timeout {
+            phase,
+            waited,
+            error,
+        };
+        self.state.record_result(result.clone());
+        result
+    }
+}
+
+fn dedupe_sorted(mut values: Vec<String>) -> Vec<String> {
+    values.sort();
+    values.dedup();
+    values
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    use serde_json::json;
+
+    #[test]
+    fn phase_display_matches_serde_name() {
+        // given
+        let phases = McpLifecyclePhase::all();
+
+        // when
+        let serialized = phases
+            .into_iter()
+            .map(|phase| {
+                (
+                    phase.to_string(),
+                    serde_json::to_value(phase).expect("serialize phase"),
+                )
+            })
+            .collect::<Vec<_>>();
+
+        // then
+        for (display, json_value) in serialized {
+            assert_eq!(json_value, json!(display));
+        }
+    }
+
+    #[test]
+    fn given_startup_path_when_running_to_cleanup_then_each_control_transition_succeeds() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+        let phases = [
+            McpLifecyclePhase::ConfigLoad,
+            McpLifecyclePhase::ServerRegistration,
+            McpLifecyclePhase::SpawnConnect,
+            McpLifecyclePhase::InitializeHandshake,
+            McpLifecyclePhase::ToolDiscovery,
+            McpLifecyclePhase::ResourceDiscovery,
+            McpLifecyclePhase::Ready,
+            McpLifecyclePhase::Invocation,
+            McpLifecyclePhase::Ready,
+            McpLifecyclePhase::Shutdown,
+            McpLifecyclePhase::Cleanup,
+        ];
+
+        // when
+        let results = phases
+            .into_iter()
+            .map(|phase| validator.run_phase(phase))
+            .collect::<Vec<_>>();
+
+        // then
+        assert!(results
+            .iter()
+            .all(|result| matches!(result, McpPhaseResult::Success { .. })));
+        assert_eq!(
+            validator.state().current_phase(),
+            Some(McpLifecyclePhase::Cleanup)
+        );
+        for phase in [
+            McpLifecyclePhase::ConfigLoad,
+            McpLifecyclePhase::ServerRegistration,
+            McpLifecyclePhase::SpawnConnect,
+            McpLifecyclePhase::InitializeHandshake,
+            McpLifecyclePhase::ToolDiscovery,
+            McpLifecyclePhase::ResourceDiscovery,
+            McpLifecyclePhase::Ready,
+            McpLifecyclePhase::Invocation,
+            McpLifecyclePhase::Shutdown,
+            McpLifecyclePhase::Cleanup,
+        ] {
+            assert!(validator.state().phase_timestamp(phase).is_some());
+        }
+    }
+
+    #[test]
+    fn given_tool_discovery_when_resource_discovery_is_skipped_then_ready_is_still_allowed() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+        for phase in [
+            McpLifecyclePhase::ConfigLoad,
+            McpLifecyclePhase::ServerRegistration,
+            McpLifecyclePhase::SpawnConnect,
+            McpLifecyclePhase::InitializeHandshake,
+            McpLifecyclePhase::ToolDiscovery,
+        ] {
+            let result = validator.run_phase(phase);
+            assert!(matches!(result, McpPhaseResult::Success { .. }));
+        }
+
+        // when
+        let result = validator.run_phase(McpLifecyclePhase::Ready);
+
+        // then
+        assert!(matches!(result, McpPhaseResult::Success { .. }));
+        assert_eq!(
+            validator.state().current_phase(),
+            Some(McpLifecyclePhase::Ready)
+        );
+    }
+
+    #[test]
+    fn validates_expected_phase_transitions() {
+        // given
+        let valid_transitions = [
+            (
+                McpLifecyclePhase::ConfigLoad,
+                McpLifecyclePhase::ServerRegistration,
+            ),
+            (
+                McpLifecyclePhase::ServerRegistration,
+                McpLifecyclePhase::SpawnConnect,
+            ),
+            (
+                McpLifecyclePhase::SpawnConnect,
+                McpLifecyclePhase::InitializeHandshake,
+            ),
+            (
+                McpLifecyclePhase::InitializeHandshake,
+                McpLifecyclePhase::ToolDiscovery,
+            ),
+            (
+                McpLifecyclePhase::ToolDiscovery,
+                McpLifecyclePhase::ResourceDiscovery,
+            ),
+            (McpLifecyclePhase::ToolDiscovery, McpLifecyclePhase::Ready),
+            (
+                McpLifecyclePhase::ResourceDiscovery,
+                McpLifecyclePhase::Ready,
+            ),
+            (McpLifecyclePhase::Ready, McpLifecyclePhase::Invocation),
+            (McpLifecyclePhase::Invocation, McpLifecyclePhase::Ready),
+            (McpLifecyclePhase::Ready, McpLifecyclePhase::Shutdown),
+            (
+                McpLifecyclePhase::Invocation,
+                McpLifecyclePhase::ErrorSurfacing,
+            ),
+            (
+                McpLifecyclePhase::ErrorSurfacing,
+                McpLifecyclePhase::Shutdown,
+            ),
+            (McpLifecyclePhase::Shutdown, McpLifecyclePhase::Cleanup),
+        ];
+
+        // when / then
+        for (from, to) in valid_transitions {
+            assert!(McpLifecycleValidator::validate_phase_transition(from, to));
+        }
+        assert!(!McpLifecycleValidator::validate_phase_transition(
+            McpLifecyclePhase::Ready,
+            McpLifecyclePhase::ConfigLoad,
+        ));
+        assert!(!McpLifecycleValidator::validate_phase_transition(
+            McpLifecyclePhase::Cleanup,
+            McpLifecyclePhase::Ready,
+        ));
+    }
+
+    #[test]
+    fn given_invalid_transition_when_running_phase_then_structured_failure_is_recorded() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+        let _ = validator.run_phase(McpLifecyclePhase::ConfigLoad);
+        let _ = validator.run_phase(McpLifecyclePhase::ServerRegistration);
+
+        // when
+        let result = validator.run_phase(McpLifecyclePhase::Ready);
+
+        // then
+        match result {
+            McpPhaseResult::Failure { phase, error } => {
+                assert_eq!(phase, McpLifecyclePhase::Ready);
+                assert!(!error.recoverable);
+                assert_eq!(error.phase, McpLifecyclePhase::Ready);
+                assert_eq!(
+                    error.context.get("from").map(String::as_str),
+                    Some("server_registration")
+                );
+                assert_eq!(error.context.get("to").map(String::as_str), Some("ready"));
+            }
+            other => panic!("expected failure result, got {other:?}"),
+        }
+        assert_eq!(
+            validator.state().current_phase(),
+            Some(McpLifecyclePhase::ErrorSurfacing)
+        );
+        assert_eq!(
+            validator
+                .state()
+                .errors_for_phase(McpLifecyclePhase::Ready)
+                .len(),
+            1
+        );
+    }
+
+    #[test]
+    fn given_each_phase_when_failure_is_recorded_then_error_is_tracked_per_phase() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+
+        // when / then
+        for phase in McpLifecyclePhase::all() {
+            let result = validator.record_failure(McpErrorSurface::new(
+                phase,
+                Some("alpha".to_string()),
+                format!("failure at {phase}"),
+                BTreeMap::from([("server".to_string(), "alpha".to_string())]),
+                phase == McpLifecyclePhase::ResourceDiscovery,
+            ));
+
+            match result {
+                McpPhaseResult::Failure {
+                    phase: failed_phase,
+                    error,
+                } => {
+                    assert_eq!(failed_phase, phase);
+                    assert_eq!(error.phase, phase);
+                    assert_eq!(
+                        error.recoverable,
+                        phase == McpLifecyclePhase::ResourceDiscovery
+                    );
+                }
+                other => panic!("expected failure result, got {other:?}"),
+            }
+            assert_eq!(validator.state().errors_for_phase(phase).len(), 1);
+        }
+    }
+
+    #[test]
+    fn given_spawn_connect_timeout_when_recorded_then_waited_duration_is_preserved() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+        let waited = Duration::from_millis(250);
+
+        // when
+        let result = validator.record_timeout(
+            McpLifecyclePhase::SpawnConnect,
+            waited,
+            Some("alpha".to_string()),
+            BTreeMap::from([("attempt".to_string(), "1".to_string())]),
+        );
+
+        // then
+        match result {
+            McpPhaseResult::Timeout {
+                phase,
+                waited: actual,
+                error,
+            } => {
+                assert_eq!(phase, McpLifecyclePhase::SpawnConnect);
+                assert_eq!(actual, waited);
+                assert!(error.recoverable);
+                assert_eq!(error.server_name.as_deref(), Some("alpha"));
+            }
+            other => panic!("expected timeout result, got {other:?}"),
+        }
+        let errors = validator
+            .state()
+            .errors_for_phase(McpLifecyclePhase::SpawnConnect);
+        assert_eq!(errors.len(), 1);
+        assert_eq!(
+            errors[0].context.get("waited_ms").map(String::as_str),
+            Some("250")
+        );
+        assert_eq!(
+            validator.state().current_phase(),
+            Some(McpLifecyclePhase::ErrorSurfacing)
+        );
+    }
+
+    #[test]
+    fn given_partial_server_health_when_building_degraded_report_then_missing_tools_are_reported() {
+        // given
+        let failed = vec![McpFailedServer {
+            server_name: "broken".to_string(),
+            phase: McpLifecyclePhase::InitializeHandshake,
+            error: McpErrorSurface::new(
+                McpLifecyclePhase::InitializeHandshake,
+                Some("broken".to_string()),
+                "initialize failed",
+                BTreeMap::from([("reason".to_string(), "broken pipe".to_string())]),
+                false,
+            ),
+        }];
+
+        // when
+        let report = McpDegradedReport::new(
+            vec!["alpha".to_string(), "beta".to_string(), "alpha".to_string()],
+            failed,
+            vec![
+                "alpha.echo".to_string(),
+                "beta.search".to_string(),
+                "alpha.echo".to_string(),
+            ],
+            vec![
+                "alpha.echo".to_string(),
+                "beta.search".to_string(),
+                "broken.fetch".to_string(),
+            ],
+        );
+
+        // then
+        assert_eq!(
+            report.working_servers,
+            vec!["alpha".to_string(), "beta".to_string()]
+        );
+        assert_eq!(report.failed_servers.len(), 1);
+        assert_eq!(report.failed_servers[0].server_name, "broken");
+        assert_eq!(
+            report.available_tools,
+            vec!["alpha.echo".to_string(), "beta.search".to_string()]
+        );
+        assert_eq!(report.missing_tools, vec!["broken.fetch".to_string()]);
+    }
+
+    #[test]
+    fn given_failure_during_resource_discovery_when_shutting_down_then_cleanup_still_succeeds() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+        for phase in [
+            McpLifecyclePhase::ConfigLoad,
+            McpLifecyclePhase::ServerRegistration,
+            McpLifecyclePhase::SpawnConnect,
+            McpLifecyclePhase::InitializeHandshake,
+            McpLifecyclePhase::ToolDiscovery,
+        ] {
+            let result = validator.run_phase(phase);
+            assert!(matches!(result, McpPhaseResult::Success { .. }));
+        }
+        let _ = validator.record_failure(McpErrorSurface::new(
+            McpLifecyclePhase::ResourceDiscovery,
+            Some("alpha".to_string()),
+            "resource listing failed",
+            BTreeMap::from([("reason".to_string(), "timeout".to_string())]),
+            true,
+        ));
+
+        // when
+        let shutdown = validator.run_phase(McpLifecyclePhase::Shutdown);
+        let cleanup = validator.run_phase(McpLifecyclePhase::Cleanup);
+
+        // then
+        assert!(matches!(shutdown, McpPhaseResult::Success { .. }));
+        assert!(matches!(cleanup, McpPhaseResult::Success { .. }));
+        assert_eq!(
+            validator.state().current_phase(),
+            Some(McpLifecyclePhase::Cleanup)
+        );
+        assert!(validator
+            .state()
+            .phase_timestamp(McpLifecyclePhase::ErrorSurfacing)
+            .is_some());
+    }
+
+    #[test]
+    fn error_surface_display_includes_phase_server_and_recoverable_flag() {
+        // given
+        let error = McpErrorSurface::new(
+            McpLifecyclePhase::SpawnConnect,
+            Some("alpha".to_string()),
+            "process exited early",
+            BTreeMap::from([("exit_code".to_string(), "1".to_string())]),
+            true,
+        );
+
+        // when
+        let rendered = error.to_string();
+
+        // then
+        assert!(rendered.contains("spawn_connect"));
+        assert!(rendered.contains("process exited early"));
+        assert!(rendered.contains("server: alpha"));
+        assert!(rendered.contains("recoverable"));
+        let trait_object: &dyn std::error::Error = &error;
+        assert_eq!(trait_object.to_string(), rendered);
+    }
+
+    #[test]
+    fn given_nonrecoverable_failure_when_returning_to_ready_then_validator_rejects_resume() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+        for phase in [
+            McpLifecyclePhase::ConfigLoad,
+            McpLifecyclePhase::ServerRegistration,
+            McpLifecyclePhase::SpawnConnect,
+            McpLifecyclePhase::InitializeHandshake,
+            McpLifecyclePhase::ToolDiscovery,
+            McpLifecyclePhase::Ready,
+        ] {
+            let result = validator.run_phase(phase);
+            assert!(matches!(result, McpPhaseResult::Success { .. }));
+        }
+        let _ = validator.record_failure(McpErrorSurface::new(
+            McpLifecyclePhase::Invocation,
+            Some("alpha".to_string()),
+            "tool call corrupted the session",
+            BTreeMap::from([("reason".to_string(), "invalid frame".to_string())]),
+            false,
+        ));
+
+        // when
+        let result = validator.run_phase(McpLifecyclePhase::Ready);
+
+        // then
+        match result {
+            McpPhaseResult::Failure { phase, error } => {
+                assert_eq!(phase, McpLifecyclePhase::Ready);
+                assert!(!error.recoverable);
+                assert!(error.message.contains("non-recoverable"));
+            }
+            other => panic!("expected failure result, got {other:?}"),
+        }
+        assert_eq!(
+            validator.state().current_phase(),
+            Some(McpLifecyclePhase::ErrorSurfacing)
+        );
+    }
+
+    #[test]
+    fn given_recoverable_failure_when_returning_to_ready_then_validator_allows_resume() {
+        // given
+        let mut validator = McpLifecycleValidator::new();
+        for phase in [
+            McpLifecyclePhase::ConfigLoad,
+            McpLifecyclePhase::ServerRegistration,
+            McpLifecyclePhase::SpawnConnect,
+            McpLifecyclePhase::InitializeHandshake,
+            McpLifecyclePhase::ToolDiscovery,
+            McpLifecyclePhase::Ready,
+        ] {
+            let result = validator.run_phase(phase);
+            assert!(matches!(result, McpPhaseResult::Success { .. }));
+        }
+        let _ = validator.record_failure(McpErrorSurface::new(
+            McpLifecyclePhase::Invocation,
+            Some("alpha".to_string()),
+            "tool call failed but can be retried",
+            BTreeMap::from([("reason".to_string(), "upstream timeout".to_string())]),
+            true,
+        ));
+
+        // when
+        let result = validator.run_phase(McpLifecyclePhase::Ready);
+
+        // then
+        assert!(matches!(result, McpPhaseResult::Success { .. }));
+        assert_eq!(
+            validator.state().current_phase(),
+            Some(McpLifecyclePhase::Ready)
+        );
+    }
+}
@@ -0,0 +1,440 @@
+//! Minimal Model Context Protocol (MCP) server.
+//!
+//! Implements a newline-safe, LSP-framed JSON-RPC server over stdio that
+//! answers `initialize`, `tools/list`, and `tools/call` requests. The framing
+//! matches the client transport implemented in [`crate::mcp_stdio`] so this
+//! server can be driven by either an external MCP client (e.g. Claude
+//! Desktop) or `claw`'s own [`McpServerManager`](crate::McpServerManager).
+//!
+//! The server is intentionally small: it exposes a list of pre-built
+//! [`McpTool`] descriptors and delegates `tools/call` to a caller-supplied
+//! handler. Tool execution itself lives in the `tools` crate; this module is
+//! purely the transport + dispatch loop.
+//!
+//! [`McpTool`]: crate::mcp_stdio::McpTool
+
+use std::io;
+
+use serde_json::{json, Value as JsonValue};
+use tokio::io::{
+    stdin, stdout, AsyncBufReadExt, AsyncReadExt, AsyncWriteExt, BufReader, Stdin, Stdout,
+};
+
+use crate::mcp_stdio::{
+    JsonRpcError, JsonRpcId, JsonRpcRequest, JsonRpcResponse, McpInitializeResult,
+    McpInitializeServerInfo, McpListToolsResult, McpTool, McpToolCallContent, McpToolCallParams,
+    McpToolCallResult,
+};
+
+/// Protocol version the server advertises during `initialize`.
+///
+/// Matches the version used by the built-in client in
+/// [`crate::mcp_stdio`], so the two stay in lockstep.
+pub const MCP_SERVER_PROTOCOL_VERSION: &str = "2025-03-26";
+
+/// Synchronous handler invoked for every `tools/call` request.
+///
+/// Returning `Ok(text)` yields a single `text` content block and
+/// `isError: false`. Returning `Err(message)` yields a `text` block with the
+/// error and `isError: true`, mirroring the error-surfacing convention used
+/// elsewhere in claw.
+pub type ToolCallHandler =
+    Box<dyn Fn(&str, &JsonValue) -> Result<String, String> + Send + Sync + 'static>;
+
+/// Configuration for an [`McpServer`] instance.
+///
+/// Named `McpServerSpec` rather than `McpServerConfig` to avoid colliding
+/// with the existing client-side [`crate::config::McpServerConfig`] that
+/// describes *remote* MCP servers the runtime connects to.
+pub struct McpServerSpec {
+    /// Name advertised in the `serverInfo` field of the `initialize` response.
+    pub server_name: String,
+    /// Version advertised in the `serverInfo` field of the `initialize`
+    /// response.
+    pub server_version: String,
+    /// Tool descriptors returned for `tools/list`.
+    pub tools: Vec<McpTool>,
+    /// Handler invoked for `tools/call`.
+    pub tool_handler: ToolCallHandler,
+}
+
+/// Minimal MCP stdio server.
+///
+/// The server runs a blocking read/dispatch/write loop over the current
+/// process's stdin/stdout, terminating cleanly when the peer closes the
+/// stream.
+pub struct McpServer {
+    spec: McpServerSpec,
+    stdin: BufReader<Stdin>,
+    stdout: Stdout,
+}
+
+impl McpServer {
+    #[must_use]
+    pub fn new(spec: McpServerSpec) -> Self {
+        Self {
+            spec,
+            stdin: BufReader::new(stdin()),
+            stdout: stdout(),
+        }
+    }
+
+    /// Runs the server until the client closes stdin.
+    ///
+    /// Returns `Ok(())` on clean EOF; any other I/O error is propagated so
+    /// callers can log and exit non-zero.
+    pub async fn run(&mut self) -> io::Result<()> {
+        loop {
+            let Some(payload) = read_frame(&mut self.stdin).await? else {
+                return Ok(());
+            };
+
+            // Requests and notifications share a wire format; the absence of
+            // `id` distinguishes notifications, which must never receive a
+            // response.
+            let message: JsonValue = match serde_json::from_slice(&payload) {
+                Ok(value) => value,
+                Err(error) => {
+                    // Parse error with null id per JSON-RPC 2.0 §4.2.
+                    let response = JsonRpcResponse::<JsonValue> {
+                        jsonrpc: "2.0".to_string(),
+                        id: JsonRpcId::Null,
+                        result: None,
+                        error: Some(JsonRpcError {
+                            code: -32700,
+                            message: format!("parse error: {error}"),
+                            data: None,
+                        }),
+                    };
+                    write_response(&mut self.stdout, &response).await?;
+                    continue;
+                }
+            };
+
+            if message.get("id").is_none() {
+                // Notification: dispatch for side effects only (e.g. log),
+                // but send no reply.
+                continue;
+            }
+
+            let request: JsonRpcRequest<JsonValue> = match serde_json::from_value(message) {
+                Ok(request) => request,
+                Err(error) => {
+                    let response = JsonRpcResponse::<JsonValue> {
+                        jsonrpc: "2.0".to_string(),
+                        id: JsonRpcId::Null,
+                        result: None,
+                        error: Some(JsonRpcError {
+                            code: -32600,
+                            message: format!("invalid request: {error}"),
+                            data: None,
+                        }),
+                    };
+                    write_response(&mut self.stdout, &response).await?;
+                    continue;
+                }
+            };
+
+            let response = self.dispatch(request);
+            write_response(&mut self.stdout, &response).await?;
+        }
+    }
+
+    fn dispatch(&self, request: JsonRpcRequest<JsonValue>) -> JsonRpcResponse<JsonValue> {
+        let id = request.id.clone();
+        match request.method.as_str() {
+            "initialize" => self.handle_initialize(id),
+            "tools/list" => self.handle_tools_list(id),
+            "tools/call" => self.handle_tools_call(id, request.params),
+            other => JsonRpcResponse {
+                jsonrpc: "2.0".to_string(),
+                id,
+                result: None,
+                error: Some(JsonRpcError {
+                    code: -32601,
+                    message: format!("method not found: {other}"),
+                    data: None,
+                }),
+            },
+        }
+    }
+
+    fn handle_initialize(&self, id: JsonRpcId) -> JsonRpcResponse<JsonValue> {
+        let result = McpInitializeResult {
+            protocol_version: MCP_SERVER_PROTOCOL_VERSION.to_string(),
+            capabilities: json!({ "tools": {} }),
+            server_info: McpInitializeServerInfo {
+                name: self.spec.server_name.clone(),
+                version: self.spec.server_version.clone(),
+            },
+        };
+        JsonRpcResponse {
+            jsonrpc: "2.0".to_string(),
+            id,
+            result: serde_json::to_value(result).ok(),
+            error: None,
+        }
+    }
+
+    fn handle_tools_list(&self, id: JsonRpcId) -> JsonRpcResponse<JsonValue> {
+        let result = McpListToolsResult {
+            tools: self.spec.tools.clone(),
+            next_cursor: None,
+        };
+        JsonRpcResponse {
+            jsonrpc: "2.0".to_string(),
+            id,
+            result: serde_json::to_value(result).ok(),
+            error: None,
+        }
+    }
+
+    fn handle_tools_call(
+        &self,
+        id: JsonRpcId,
+        params: Option<JsonValue>,
+    ) -> JsonRpcResponse<JsonValue> {
+        let Some(params) = params else {
+            return invalid_params_response(id, "missing params for tools/call");
+        };
+        let call: McpToolCallParams = match serde_json::from_value(params) {
+            Ok(value) => value,
+            Err(error) => {
+                return invalid_params_response(id, &format!("invalid tools/call params: {error}"));
+            }
+        };
+        let arguments = call.arguments.unwrap_or_else(|| json!({}));
+        let tool_result = (self.spec.tool_handler)(&call.name, &arguments);
+        let (text, is_error) = match tool_result {
+            Ok(text) => (text, false),
+            Err(message) => (message, true),
+        };
+        let mut data = std::collections::BTreeMap::new();
+        data.insert("text".to_string(), JsonValue::String(text));
+        let call_result = McpToolCallResult {
+            content: vec![McpToolCallContent {
+                kind: "text".to_string(),
+                data,
+            }],
+            structured_content: None,
+            is_error: Some(is_error),
+            meta: None,
+        };
+        JsonRpcResponse {
+            jsonrpc: "2.0".to_string(),
+            id,
+            result: serde_json::to_value(call_result).ok(),
+            error: None,
+        }
+    }
+}
+
+fn invalid_params_response(id: JsonRpcId, message: &str) -> JsonRpcResponse<JsonValue> {
+    JsonRpcResponse {
+        jsonrpc: "2.0".to_string(),
+        id,
+        result: None,
+        error: Some(JsonRpcError {
+            code: -32602,
+            message: message.to_string(),
+            data: None,
+        }),
+    }
+}
+
+/// Reads a single LSP-framed JSON-RPC payload from `reader`.
+///
+/// Returns `Ok(None)` on clean EOF before any header bytes have been read,
+/// matching how [`crate::mcp_stdio::McpStdioProcess`] treats stream closure.
+async fn read_frame(reader: &mut BufReader<Stdin>) -> io::Result<Option<Vec<u8>>> {
+    let mut content_length: Option<usize> = None;
+    let mut first_header = true;
+    loop {
+        let mut line = String::new();
+        let bytes_read = reader.read_line(&mut line).await?;
+        if bytes_read == 0 {
+            if first_header {
+                return Ok(None);
+            }
+            return Err(io::Error::new(
+                io::ErrorKind::UnexpectedEof,
+                "MCP stdio stream closed while reading headers",
+            ));
+        }
+        first_header = false;
+        if line == "\r\n" || line == "\n" {
+            break;
+        }
+        let header = line.trim_end_matches(['\r', '\n']);
+        if let Some((name, value)) = header.split_once(':') {
+            if name.trim().eq_ignore_ascii_case("Content-Length") {
+                let parsed = value
+                    .trim()
+                    .parse::<usize>()
+                    .map_err(|error| io::Error::new(io::ErrorKind::InvalidData, error))?;
+                content_length = Some(parsed);
+            }
+        }
+    }
+
+    let content_length = content_length.ok_or_else(|| {
+        io::Error::new(io::ErrorKind::InvalidData, "missing Content-Length header")
+    })?;
+    let mut payload = vec![0_u8; content_length];
+    reader.read_exact(&mut payload).await?;
+    Ok(Some(payload))
+}
+
+async fn write_response(
+    stdout: &mut Stdout,
+    response: &JsonRpcResponse<JsonValue>,
+) -> io::Result<()> {
+    let body = serde_json::to_vec(response)
+        .map_err(|error| io::Error::new(io::ErrorKind::InvalidData, error))?;
+    let header = format!("Content-Length: {}\r\n\r\n", body.len());
+    stdout.write_all(header.as_bytes()).await?;
+    stdout.write_all(&body).await?;
+    stdout.flush().await
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn dispatch_initialize_returns_server_info() {
+        let server = McpServer {
+            spec: McpServerSpec {
+                server_name: "test".to_string(),
+                server_version: "9.9.9".to_string(),
+                tools: Vec::new(),
+                tool_handler: Box::new(|_, _| Ok(String::new())),
+            },
+            stdin: BufReader::new(stdin()),
+            stdout: stdout(),
+        };
+        let request = JsonRpcRequest::<JsonValue> {
+            jsonrpc: "2.0".to_string(),
+            id: JsonRpcId::Number(1),
+            method: "initialize".to_string(),
+            params: None,
+        };
+        let response = server.dispatch(request);
+        assert_eq!(response.id, JsonRpcId::Number(1));
+        assert!(response.error.is_none());
+        let result = response.result.expect("initialize result");
+        assert_eq!(result["protocolVersion"], MCP_SERVER_PROTOCOL_VERSION);
+        assert_eq!(result["serverInfo"]["name"], "test");
+        assert_eq!(result["serverInfo"]["version"], "9.9.9");
+    }
+
+    #[test]
+    fn dispatch_tools_list_returns_registered_tools() {
+        let tool = McpTool {
+            name: "echo".to_string(),
+            description: Some("Echo".to_string()),
+            input_schema: Some(json!({"type": "object"})),
+            annotations: None,
+            meta: None,
+        };
+        let server = McpServer {
+            spec: McpServerSpec {
+                server_name: "test".to_string(),
+                server_version: "0.0.0".to_string(),
+                tools: vec![tool.clone()],
+                tool_handler: Box::new(|_, _| Ok(String::new())),
+            },
+            stdin: BufReader::new(stdin()),
+            stdout: stdout(),
+        };
+        let request = JsonRpcRequest::<JsonValue> {
+            jsonrpc: "2.0".to_string(),
+            id: JsonRpcId::Number(2),
+            method: "tools/list".to_string(),
+            params: None,
+        };
+        let response = server.dispatch(request);
+        assert!(response.error.is_none());
+        let result = response.result.expect("tools/list result");
+        assert_eq!(result["tools"][0]["name"], "echo");
+    }
+
+    #[test]
+    fn dispatch_tools_call_wraps_handler_output() {
+        let server = McpServer {
+            spec: McpServerSpec {
+                server_name: "test".to_string(),
+                server_version: "0.0.0".to_string(),
+                tools: Vec::new(),
+                tool_handler: Box::new(|name, args| Ok(format!("called {name} with {args}"))),
+            },
+            stdin: BufReader::new(stdin()),
+            stdout: stdout(),
+        };
+        let request = JsonRpcRequest::<JsonValue> {
+            jsonrpc: "2.0".to_string(),
+            id: JsonRpcId::Number(3),
+            method: "tools/call".to_string(),
+            params: Some(json!({
+                "name": "echo",
+                "arguments": {"text": "hi"}
+            })),
+        };
+        let response = server.dispatch(request);
+        assert!(response.error.is_none());
+        let result = response.result.expect("tools/call result");
+        assert_eq!(result["isError"], false);
+        assert_eq!(result["content"][0]["type"], "text");
+        assert!(result["content"][0]["text"]
+            .as_str()
+            .unwrap()
+            .starts_with("called echo"));
+    }
+
+    #[test]
+    fn dispatch_tools_call_surfaces_handler_error() {
+        let server = McpServer {
+            spec: McpServerSpec {
+                server_name: "test".to_string(),
+                server_version: "0.0.0".to_string(),
+                tools: Vec::new(),
+                tool_handler: Box::new(|_, _| Err("boom".to_string())),
+            },
+            stdin: BufReader::new(stdin()),
+            stdout: stdout(),
+        };
+        let request = JsonRpcRequest::<JsonValue> {
+            jsonrpc: "2.0".to_string(),
+            id: JsonRpcId::Number(4),
+            method: "tools/call".to_string(),
+            params: Some(json!({"name": "broken"})),
+        };
+        let response = server.dispatch(request);
+        let result = response.result.expect("tools/call result");
+        assert_eq!(result["isError"], true);
+        assert_eq!(result["content"][0]["text"], "boom");
+    }
+
+    #[test]
+    fn dispatch_unknown_method_returns_method_not_found() {
+        let server = McpServer {
+            spec: McpServerSpec {
+                server_name: "test".to_string(),
+                server_version: "0.0.0".to_string(),
+                tools: Vec::new(),
+                tool_handler: Box::new(|_, _| Ok(String::new())),
+            },
+            stdin: BufReader::new(stdin()),
+            stdout: stdout(),
+        };
+        let request = JsonRpcRequest::<JsonValue> {
+            jsonrpc: "2.0".to_string(),
+            id: JsonRpcId::Number(5),
+            method: "nonsense".to_string(),
+            params: None,
+        };
+        let response = server.dispatch(request);
+        let error = response.error.expect("error payload");
+        assert_eq!(error.code, -32601);
+    }
+}
@@ -0,0 +1,920 @@
+#![allow(
+    clippy::await_holding_lock,
+    clippy::doc_markdown,
+    clippy::match_same_arms,
+    clippy::must_use_candidate,
+    clippy::uninlined_format_args,
+    clippy::unnested_or_patterns
+)]
+//! Bridge between MCP tool surface (ListMcpResources, ReadMcpResource, McpAuth, MCP)
+//! and the existing McpServerManager runtime.
+//!
+//! Provides a stateful client registry that tool handlers can use to
+//! connect to MCP servers and invoke their capabilities.
+
+use std::collections::HashMap;
+use std::sync::{Arc, Mutex, OnceLock};
+
+use crate::mcp::mcp_tool_name;
+use crate::mcp_stdio::McpServerManager;
+use serde::{Deserialize, Serialize};
+
+/// Status of a managed MCP server connection.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum McpConnectionStatus {
+    Disconnected,
+    Connecting,
+    Connected,
+    AuthRequired,
+    Error,
+}
+
+impl std::fmt::Display for McpConnectionStatus {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Disconnected => write!(f, "disconnected"),
+            Self::Connecting => write!(f, "connecting"),
+            Self::Connected => write!(f, "connected"),
+            Self::AuthRequired => write!(f, "auth_required"),
+            Self::Error => write!(f, "error"),
+        }
+    }
+}
+
+/// Metadata about an MCP resource.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct McpResourceInfo {
+    pub uri: String,
+    pub name: String,
+    pub description: Option<String>,
+    pub mime_type: Option<String>,
+}
+
+/// Metadata about an MCP tool exposed by a server.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct McpToolInfo {
+    pub name: String,
+    pub description: Option<String>,
+    pub input_schema: Option<serde_json::Value>,
+}
+
+/// Tracked state of an MCP server connection.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct McpServerState {
+    pub server_name: String,
+    pub status: McpConnectionStatus,
+    pub tools: Vec<McpToolInfo>,
+    pub resources: Vec<McpResourceInfo>,
+    pub server_info: Option<String>,
+    pub error_message: Option<String>,
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct McpToolRegistry {
+    inner: Arc<Mutex<HashMap<String, McpServerState>>>,
+    manager: Arc<OnceLock<Arc<Mutex<McpServerManager>>>>,
+}
+
+impl McpToolRegistry {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    pub fn set_manager(
+        &self,
+        manager: Arc<Mutex<McpServerManager>>,
+    ) -> Result<(), Arc<Mutex<McpServerManager>>> {
+        self.manager.set(manager)
+    }
+
+    pub fn register_server(
+        &self,
+        server_name: &str,
+        status: McpConnectionStatus,
+        tools: Vec<McpToolInfo>,
+        resources: Vec<McpResourceInfo>,
+        server_info: Option<String>,
+    ) {
+        let mut inner = self.inner.lock().expect("mcp registry lock poisoned");
+        inner.insert(
+            server_name.to_owned(),
+            McpServerState {
+                server_name: server_name.to_owned(),
+                status,
+                tools,
+                resources,
+                server_info,
+                error_message: None,
+            },
+        );
+    }
+
+    pub fn get_server(&self, server_name: &str) -> Option<McpServerState> {
+        let inner = self.inner.lock().expect("mcp registry lock poisoned");
+        inner.get(server_name).cloned()
+    }
+
+    pub fn list_servers(&self) -> Vec<McpServerState> {
+        let inner = self.inner.lock().expect("mcp registry lock poisoned");
+        inner.values().cloned().collect()
+    }
+
+    pub fn list_resources(&self, server_name: &str) -> Result<Vec<McpResourceInfo>, String> {
+        let inner = self.inner.lock().expect("mcp registry lock poisoned");
+        match inner.get(server_name) {
+            Some(state) => {
+                if state.status != McpConnectionStatus::Connected {
+                    return Err(format!(
+                        "server '{}' is not connected (status: {})",
+                        server_name, state.status
+                    ));
+                }
+                Ok(state.resources.clone())
+            }
+            None => Err(format!("server '{}' not found", server_name)),
+        }
+    }
+
+    pub fn read_resource(&self, server_name: &str, uri: &str) -> Result<McpResourceInfo, String> {
+        let inner = self.inner.lock().expect("mcp registry lock poisoned");
+        let state = inner
+            .get(server_name)
+            .ok_or_else(|| format!("server '{}' not found", server_name))?;
+
+        if state.status != McpConnectionStatus::Connected {
+            return Err(format!(
+                "server '{}' is not connected (status: {})",
+                server_name, state.status
+            ));
+        }
+
+        state
+            .resources
+            .iter()
+            .find(|r| r.uri == uri)
+            .cloned()
+            .ok_or_else(|| format!("resource '{}' not found on server '{}'", uri, server_name))
+    }
+
+    pub fn list_tools(&self, server_name: &str) -> Result<Vec<McpToolInfo>, String> {
+        let inner = self.inner.lock().expect("mcp registry lock poisoned");
+        match inner.get(server_name) {
+            Some(state) => {
+                if state.status != McpConnectionStatus::Connected {
+                    return Err(format!(
+                        "server '{}' is not connected (status: {})",
+                        server_name, state.status
+                    ));
+                }
+                Ok(state.tools.clone())
+            }
+            None => Err(format!("server '{}' not found", server_name)),
+        }
+    }
+
+    fn spawn_tool_call(
+        manager: Arc<Mutex<McpServerManager>>,
+        qualified_tool_name: String,
+        arguments: Option<serde_json::Value>,
+    ) -> Result<serde_json::Value, String> {
+        let join_handle = std::thread::Builder::new()
+            .name(format!("mcp-tool-call-{qualified_tool_name}"))
+            .spawn(move || {
+                let runtime = tokio::runtime::Builder::new_current_thread()
+                    .enable_all()
+                    .build()
+                    .map_err(|error| format!("failed to create MCP tool runtime: {error}"))?;
+
+                runtime.block_on(async move {
+                    let response = {
+                        let mut manager = manager
+                            .lock()
+                            .map_err(|_| "mcp server manager lock poisoned".to_string())?;
+                        manager
+                            .discover_tools()
+                            .await
+                            .map_err(|error| error.to_string())?;
+                        let response = manager
+                            .call_tool(&qualified_tool_name, arguments)
+                            .await
+                            .map_err(|error| error.to_string());
+                        let shutdown = manager.shutdown().await.map_err(|error| error.to_string());
+
+                        match (response, shutdown) {
+                            (Ok(response), Ok(())) => Ok(response),
+                            (Err(error), Ok(())) | (Err(error), Err(_)) => Err(error),
+                            (Ok(_), Err(error)) => Err(error),
+                        }
+                    }?;
+
+                    if let Some(error) = response.error {
+                        return Err(format!(
+                            "MCP server returned JSON-RPC error for tools/call: {} ({})",
+                            error.message, error.code
+                        ));
+                    }
+
+                    let result = response.result.ok_or_else(|| {
+                        "MCP server returned no result for tools/call".to_string()
+                    })?;
+
+                    serde_json::to_value(result)
+                        .map_err(|error| format!("failed to serialize MCP tool result: {error}"))
+                })
+            })
+            .map_err(|error| format!("failed to spawn MCP tool call thread: {error}"))?;
+
+        join_handle.join().map_err(|panic_payload| {
+            if let Some(message) = panic_payload.downcast_ref::<&str>() {
+                format!("MCP tool call thread panicked: {message}")
+            } else if let Some(message) = panic_payload.downcast_ref::<String>() {
+                format!("MCP tool call thread panicked: {message}")
+            } else {
+                "MCP tool call thread panicked".to_string()
+            }
+        })?
+    }
+
+    pub fn call_tool(
+        &self,
+        server_name: &str,
+        tool_name: &str,
+        arguments: &serde_json::Value,
+    ) -> Result<serde_json::Value, String> {
+        let inner = self.inner.lock().expect("mcp registry lock poisoned");
+        let state = inner
+            .get(server_name)
+            .ok_or_else(|| format!("server '{}' not found", server_name))?;
+
+        if state.status != McpConnectionStatus::Connected {
+            return Err(format!(
+                "server '{}' is not connected (status: {})",
+                server_name, state.status
+            ));
+        }
+
+        if !state.tools.iter().any(|t| t.name == tool_name) {
+            return Err(format!(
+                "tool '{}' not found on server '{}'",
+                tool_name, server_name
+            ));
+        }
+
+        drop(inner);
+
+        let manager = self
+            .manager
+            .get()
+            .cloned()
+            .ok_or_else(|| "MCP server manager is not configured".to_string())?;
+
+        Self::spawn_tool_call(
+            manager,
+            mcp_tool_name(server_name, tool_name),
+            (!arguments.is_null()).then(|| arguments.clone()),
+        )
+    }
+
+    /// Set auth status for a server.
+    pub fn set_auth_status(
+        &self,
+        server_name: &str,
+        status: McpConnectionStatus,
+    ) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("mcp registry lock poisoned");
+        let state = inner
+            .get_mut(server_name)
+            .ok_or_else(|| format!("server '{}' not found", server_name))?;
+        state.status = status;
+        Ok(())
+    }
+
+    /// Disconnect / remove a server.
+    pub fn disconnect(&self, server_name: &str) -> Option<McpServerState> {
+        let mut inner = self.inner.lock().expect("mcp registry lock poisoned");
+        inner.remove(server_name)
+    }
+
+    /// Number of registered servers.
+    #[must_use]
+    pub fn len(&self) -> usize {
+        let inner = self.inner.lock().expect("mcp registry lock poisoned");
+        inner.len()
+    }
+
+    #[must_use]
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+}
+
+#[cfg(all(test, unix))]
+mod tests {
+    use std::collections::BTreeMap;
+    use std::fs;
+    use std::os::unix::fs::PermissionsExt;
+    use std::path::{Path, PathBuf};
+    use std::sync::atomic::{AtomicU64, Ordering};
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    use super::*;
+    use crate::config::{
+        ConfigSource, McpServerConfig, McpStdioServerConfig, ScopedMcpServerConfig,
+    };
+
+    fn temp_dir() -> PathBuf {
+        static NEXT_TEMP_DIR_ID: AtomicU64 = AtomicU64::new(0);
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_nanos();
+        let unique_id = NEXT_TEMP_DIR_ID.fetch_add(1, Ordering::Relaxed);
+        std::env::temp_dir().join(format!("runtime-mcp-tool-bridge-{nanos}-{unique_id}"))
+    }
+
+    fn cleanup_script(script_path: &Path) {
+        if let Some(root) = script_path.parent() {
+            let _ = fs::remove_dir_all(root);
+        }
+    }
+
+    fn write_bridge_mcp_server_script() -> PathBuf {
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("temp dir");
+        let script_path = root.join("bridge-mcp-server.py");
+        let script = [
+            "#!/usr/bin/env python3",
+            "import json, os, sys",
+            "LABEL = os.environ.get('MCP_SERVER_LABEL', 'server')",
+            "LOG_PATH = os.environ.get('MCP_LOG_PATH')",
+            "",
+            "def log(method):",
+            "    if LOG_PATH:",
+            "        with open(LOG_PATH, 'a', encoding='utf-8') as handle:",
+            "            handle.write(f'{method}\\n')",
+            "",
+            "def read_message():",
+            "    header = b''",
+            r"    while not header.endswith(b'\r\n\r\n'):",
+            "        chunk = sys.stdin.buffer.read(1)",
+            "        if not chunk:",
+            "            return None",
+            "        header += chunk",
+            "    length = 0",
+            r"    for line in header.decode().split('\r\n'):",
+            r"        if line.lower().startswith('content-length:'):",
+            r"            length = int(line.split(':', 1)[1].strip())",
+            "    payload = sys.stdin.buffer.read(length)",
+            "    return json.loads(payload.decode())",
+            "",
+            "def send_message(message):",
+            "    payload = json.dumps(message).encode()",
+            r"    sys.stdout.buffer.write(f'Content-Length: {len(payload)}\r\n\r\n'.encode() + payload)",
+            "    sys.stdout.buffer.flush()",
+            "",
+            "while True:",
+            "    request = read_message()",
+            "    if request is None:",
+            "        break",
+            "    method = request['method']",
+            "    log(method)",
+            "    if method == 'initialize':",
+            "        send_message({",
+            "            'jsonrpc': '2.0',",
+            "            'id': request['id'],",
+            "            'result': {",
+            "                'protocolVersion': request['params']['protocolVersion'],",
+            "                'capabilities': {'tools': {}},",
+            "                'serverInfo': {'name': LABEL, 'version': '1.0.0'}",
+            "            }",
+            "        })",
+            "    elif method == 'tools/list':",
+            "        send_message({",
+            "            'jsonrpc': '2.0',",
+            "            'id': request['id'],",
+            "            'result': {",
+            "                'tools': [",
+            "                    {",
+            "                        'name': 'echo',",
+            "                        'description': f'Echo tool for {LABEL}',",
+            "                        'inputSchema': {",
+            "                            'type': 'object',",
+            "                            'properties': {'text': {'type': 'string'}},",
+            "                            'required': ['text']",
+            "                        }",
+            "                    }",
+            "                ]",
+            "            }",
+            "        })",
+            "    elif method == 'tools/call':",
+            "        args = request['params'].get('arguments') or {}",
+            "        text = args.get('text', '')",
+            "        send_message({",
+            "            'jsonrpc': '2.0',",
+            "            'id': request['id'],",
+            "            'result': {",
+            "                'content': [{'type': 'text', 'text': f'{LABEL}:{text}'}],",
+            "                'structuredContent': {'server': LABEL, 'echoed': text},",
+            "                'isError': False",
+            "            }",
+            "        })",
+            "    else:",
+            "        send_message({",
+            "            'jsonrpc': '2.0',",
+            "            'id': request['id'],",
+            "            'error': {'code': -32601, 'message': f'unknown method: {method}'},",
+            "        })",
+            "",
+        ]
+        .join("\n");
+        fs::write(&script_path, script).expect("write script");
+        let mut permissions = fs::metadata(&script_path).expect("metadata").permissions();
+        permissions.set_mode(0o755);
+        fs::set_permissions(&script_path, permissions).expect("chmod");
+        script_path
+    }
+
+    fn manager_server_config(
+        script_path: &Path,
+        server_name: &str,
+        log_path: &Path,
+    ) -> ScopedMcpServerConfig {
+        ScopedMcpServerConfig {
+            scope: ConfigSource::Local,
+            config: McpServerConfig::Stdio(McpStdioServerConfig {
+                command: "python3".to_string(),
+                args: vec![script_path.to_string_lossy().into_owned()],
+                env: BTreeMap::from([
+                    ("MCP_SERVER_LABEL".to_string(), server_name.to_string()),
+                    (
+                        "MCP_LOG_PATH".to_string(),
+                        log_path.to_string_lossy().into_owned(),
+                    ),
+                ]),
+                tool_call_timeout_ms: Some(1_000),
+            }),
+        }
+    }
+
+    #[test]
+    fn registers_and_retrieves_server() {
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "test-server",
+            McpConnectionStatus::Connected,
+            vec![McpToolInfo {
+                name: "greet".into(),
+                description: Some("Greet someone".into()),
+                input_schema: None,
+            }],
+            vec![McpResourceInfo {
+                uri: "res://data".into(),
+                name: "Data".into(),
+                description: None,
+                mime_type: Some("application/json".into()),
+            }],
+            Some("TestServer v1.0".into()),
+        );
+
+        let server = registry.get_server("test-server").expect("should exist");
+        assert_eq!(server.status, McpConnectionStatus::Connected);
+        assert_eq!(server.tools.len(), 1);
+        assert_eq!(server.resources.len(), 1);
+    }
+
+    #[test]
+    fn lists_resources_from_connected_server() {
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::Connected,
+            vec![],
+            vec![McpResourceInfo {
+                uri: "res://alpha".into(),
+                name: "Alpha".into(),
+                description: None,
+                mime_type: None,
+            }],
+            None,
+        );
+
+        let resources = registry.list_resources("srv").expect("should succeed");
+        assert_eq!(resources.len(), 1);
+        assert_eq!(resources[0].uri, "res://alpha");
+    }
+
+    #[test]
+    fn rejects_resource_listing_for_disconnected_server() {
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::Disconnected,
+            vec![],
+            vec![],
+            None,
+        );
+        assert!(registry.list_resources("srv").is_err());
+    }
+
+    #[test]
+    fn reads_specific_resource() {
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::Connected,
+            vec![],
+            vec![McpResourceInfo {
+                uri: "res://data".into(),
+                name: "Data".into(),
+                description: Some("Test data".into()),
+                mime_type: Some("text/plain".into()),
+            }],
+            None,
+        );
+
+        let resource = registry
+            .read_resource("srv", "res://data")
+            .expect("should find");
+        assert_eq!(resource.name, "Data");
+
+        assert!(registry.read_resource("srv", "res://missing").is_err());
+    }
+
+    #[test]
+    fn given_connected_server_without_manager_when_calling_tool_then_it_errors() {
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::Connected,
+            vec![McpToolInfo {
+                name: "greet".into(),
+                description: None,
+                input_schema: None,
+            }],
+            vec![],
+            None,
+        );
+
+        let error = registry
+            .call_tool("srv", "greet", &serde_json::json!({"name": "world"}))
+            .expect_err("should require a configured manager");
+        assert!(error.contains("MCP server manager is not configured"));
+
+        // Unknown tool should fail
+        assert!(registry
+            .call_tool("srv", "missing", &serde_json::json!({}))
+            .is_err());
+    }
+
+    #[test]
+    fn given_connected_server_with_manager_when_calling_tool_then_it_returns_live_result() {
+        let script_path = write_bridge_mcp_server_script();
+        let root = script_path.parent().expect("script parent");
+        let log_path = root.join("bridge.log");
+        let servers = BTreeMap::from([(
+            "alpha".to_string(),
+            manager_server_config(&script_path, "alpha", &log_path),
+        )]);
+        let manager = Arc::new(Mutex::new(McpServerManager::from_servers(&servers)));
+
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "alpha",
+            McpConnectionStatus::Connected,
+            vec![McpToolInfo {
+                name: "echo".into(),
+                description: Some("Echo tool for alpha".into()),
+                input_schema: Some(serde_json::json!({
+                    "type": "object",
+                    "properties": {"text": {"type": "string"}},
+                    "required": ["text"]
+                })),
+            }],
+            vec![],
+            Some("bridge test server".into()),
+        );
+        registry
+            .set_manager(Arc::clone(&manager))
+            .expect("manager should only be set once");
+
+        let result = registry
+            .call_tool("alpha", "echo", &serde_json::json!({"text": "hello"}))
+            .expect("should return live MCP result");
+
+        assert_eq!(
+            result["structuredContent"]["server"],
+            serde_json::json!("alpha")
+        );
+        assert_eq!(
+            result["structuredContent"]["echoed"],
+            serde_json::json!("hello")
+        );
+        assert_eq!(
+            result["content"][0]["text"],
+            serde_json::json!("alpha:hello")
+        );
+
+        let log = fs::read_to_string(&log_path).expect("read log");
+        assert_eq!(
+            log.lines().collect::<Vec<_>>(),
+            vec!["initialize", "tools/list", "tools/call"]
+        );
+
+        cleanup_script(&script_path);
+    }
+
+    #[test]
+    fn rejects_tool_call_on_disconnected_server() {
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::AuthRequired,
+            vec![McpToolInfo {
+                name: "greet".into(),
+                description: None,
+                input_schema: None,
+            }],
+            vec![],
+            None,
+        );
+
+        assert!(registry
+            .call_tool("srv", "greet", &serde_json::json!({}))
+            .is_err());
+    }
+
+    #[test]
+    fn sets_auth_and_disconnects() {
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::AuthRequired,
+            vec![],
+            vec![],
+            None,
+        );
+
+        registry
+            .set_auth_status("srv", McpConnectionStatus::Connected)
+            .expect("should succeed");
+        let state = registry.get_server("srv").unwrap();
+        assert_eq!(state.status, McpConnectionStatus::Connected);
+
+        let removed = registry.disconnect("srv");
+        assert!(removed.is_some());
+        assert!(registry.is_empty());
+    }
+
+    #[test]
+    fn rejects_operations_on_missing_server() {
+        let registry = McpToolRegistry::new();
+        assert!(registry.list_resources("missing").is_err());
+        assert!(registry.read_resource("missing", "uri").is_err());
+        assert!(registry.list_tools("missing").is_err());
+        assert!(registry
+            .call_tool("missing", "tool", &serde_json::json!({}))
+            .is_err());
+        assert!(registry
+            .set_auth_status("missing", McpConnectionStatus::Connected)
+            .is_err());
+    }
+
+    #[test]
+    fn mcp_connection_status_display_all_variants() {
+        // given
+        let cases = [
+            (McpConnectionStatus::Disconnected, "disconnected"),
+            (McpConnectionStatus::Connecting, "connecting"),
+            (McpConnectionStatus::Connected, "connected"),
+            (McpConnectionStatus::AuthRequired, "auth_required"),
+            (McpConnectionStatus::Error, "error"),
+        ];
+
+        // when
+        let rendered: Vec<_> = cases
+            .into_iter()
+            .map(|(status, expected)| (status.to_string(), expected))
+            .collect();
+
+        // then
+        assert_eq!(
+            rendered,
+            vec![
+                ("disconnected".to_string(), "disconnected"),
+                ("connecting".to_string(), "connecting"),
+                ("connected".to_string(), "connected"),
+                ("auth_required".to_string(), "auth_required"),
+                ("error".to_string(), "error"),
+            ]
+        );
+    }
+
+    #[test]
+    fn list_servers_returns_all_registered() {
+        // given
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "alpha",
+            McpConnectionStatus::Connected,
+            vec![],
+            vec![],
+            None,
+        );
+        registry.register_server(
+            "beta",
+            McpConnectionStatus::Connecting,
+            vec![],
+            vec![],
+            None,
+        );
+
+        // when
+        let servers = registry.list_servers();
+
+        // then
+        assert_eq!(servers.len(), 2);
+        assert!(servers.iter().any(|server| server.server_name == "alpha"));
+        assert!(servers.iter().any(|server| server.server_name == "beta"));
+    }
+
+    #[test]
+    fn list_tools_from_connected_server() {
+        // given
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::Connected,
+            vec![McpToolInfo {
+                name: "inspect".into(),
+                description: Some("Inspect data".into()),
+                input_schema: Some(serde_json::json!({"type": "object"})),
+            }],
+            vec![],
+            None,
+        );
+
+        // when
+        let tools = registry.list_tools("srv").expect("tools should list");
+
+        // then
+        assert_eq!(tools.len(), 1);
+        assert_eq!(tools[0].name, "inspect");
+    }
+
+    #[test]
+    fn list_tools_rejects_disconnected_server() {
+        // given
+        let registry = McpToolRegistry::new();
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::AuthRequired,
+            vec![],
+            vec![],
+            None,
+        );
+
+        // when
+        let result = registry.list_tools("srv");
+
+        // then
+        let error = result.expect_err("non-connected server should fail");
+        assert!(error.contains("not connected"));
+        assert!(error.contains("auth_required"));
+    }
+
+    #[test]
+    fn list_tools_rejects_missing_server() {
+        // given
+        let registry = McpToolRegistry::new();
+
+        // when
+        let result = registry.list_tools("missing");
+
+        // then
+        assert_eq!(
+            result.expect_err("missing server should fail"),
+            "server 'missing' not found"
+        );
+    }
+
+    #[test]
+    fn get_server_returns_none_for_missing() {
+        // given
+        let registry = McpToolRegistry::new();
+
+        // when
+        let server = registry.get_server("missing");
+
+        // then
+        assert!(server.is_none());
+    }
+
+    #[test]
+    fn call_tool_payload_structure() {
+        let script_path = write_bridge_mcp_server_script();
+        let root = script_path.parent().expect("script parent");
+        let log_path = root.join("payload.log");
+        let servers = BTreeMap::from([(
+            "srv".to_string(),
+            manager_server_config(&script_path, "srv", &log_path),
+        )]);
+        let registry = McpToolRegistry::new();
+        let arguments = serde_json::json!({"text": "world"});
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::Connected,
+            vec![McpToolInfo {
+                name: "echo".into(),
+                description: Some("Echo tool for srv".into()),
+                input_schema: Some(serde_json::json!({
+                    "type": "object",
+                    "properties": {"text": {"type": "string"}},
+                    "required": ["text"]
+                })),
+            }],
+            vec![],
+            None,
+        );
+        registry
+            .set_manager(Arc::new(Mutex::new(McpServerManager::from_servers(
+                &servers,
+            ))))
+            .expect("manager should only be set once");
+
+        let result = registry
+            .call_tool("srv", "echo", &arguments)
+            .expect("tool should return live payload");
+
+        assert_eq!(result["structuredContent"]["server"], "srv");
+        assert_eq!(result["structuredContent"]["echoed"], "world");
+        assert_eq!(result["content"][0]["text"], "srv:world");
+
+        cleanup_script(&script_path);
+    }
+
+    #[test]
+    fn upsert_overwrites_existing_server() {
+        // given
+        let registry = McpToolRegistry::new();
+        registry.register_server("srv", McpConnectionStatus::Connecting, vec![], vec![], None);
+
+        // when
+        registry.register_server(
+            "srv",
+            McpConnectionStatus::Connected,
+            vec![McpToolInfo {
+                name: "inspect".into(),
+                description: None,
+                input_schema: None,
+            }],
+            vec![],
+            Some("Inspector".into()),
+        );
+        let state = registry.get_server("srv").expect("server should exist");
+
+        // then
+        assert_eq!(state.status, McpConnectionStatus::Connected);
+        assert_eq!(state.tools.len(), 1);
+        assert_eq!(state.server_info.as_deref(), Some("Inspector"));
+    }
+
+    #[test]
+    fn disconnect_missing_returns_none() {
+        // given
+        let registry = McpToolRegistry::new();
+
+        // when
+        let removed = registry.disconnect("missing");
+
+        // then
+        assert!(removed.is_none());
+    }
+
+    #[test]
+    fn len_and_is_empty_transitions() {
+        // given
+        let registry = McpToolRegistry::new();
+
+        // when
+        registry.register_server(
+            "alpha",
+            McpConnectionStatus::Connected,
+            vec![],
+            vec![],
+            None,
+        );
+        registry.register_server("beta", McpConnectionStatus::Connected, vec![], vec![], None);
+        let after_create = registry.len();
+        registry.disconnect("alpha");
+        let after_first_remove = registry.len();
+        registry.disconnect("beta");
+
+        // then
+        assert_eq!(after_create, 2);
+        assert_eq!(after_first_remove, 1);
+        assert_eq!(registry.len(), 0);
+        assert!(registry.is_empty());
+    }
+}
@@ -9,6 +9,7 @@ use sha2::{Digest, Sha256};

 use crate::config::OAuthConfig;

+/// Persisted OAuth access token bundle used by the CLI.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub struct OAuthTokenSet {
    pub access_token: String,
@@ -17,6 +18,7 @@ pub struct OAuthTokenSet {
    pub scopes: Vec<String>,
 }

+/// PKCE verifier/challenge pair generated for an OAuth authorization flow.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct PkceCodePair {
    pub verifier: String,
@@ -24,6 +26,7 @@ pub struct PkceCodePair {
    pub challenge_method: PkceChallengeMethod,
 }

+/// Challenge algorithms supported by the local PKCE helpers.
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub enum PkceChallengeMethod {
    S256,
@@ -38,6 +41,7 @@ impl PkceChallengeMethod {
    }
 }

+/// Parameters needed to build an authorization URL for browser-based login.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct OAuthAuthorizationRequest {
    pub authorize_url: String,
@@ -50,6 +54,7 @@ pub struct OAuthAuthorizationRequest {
    pub extra_params: BTreeMap<String, String>,
 }

+/// Request body for exchanging an OAuth authorization code for tokens.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct OAuthTokenExchangeRequest {
    pub grant_type: &'static str,
@@ -60,6 +65,7 @@ pub struct OAuthTokenExchangeRequest {
    pub state: String,
 }

+/// Request body for refreshing an existing OAuth token set.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct OAuthRefreshRequest {
    pub grant_type: &'static str,
@@ -68,6 +74,7 @@ pub struct OAuthRefreshRequest {
    pub scopes: Vec<String>,
 }

+/// Parsed query parameters returned to the local OAuth callback endpoint.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct OAuthCallbackParams {
    pub code: Option<String>,
@@ -327,15 +334,16 @@ fn credentials_home_dir() -> io::Result<PathBuf> {
    if let Some(path) = std::env::var_os("CLAW_CONFIG_HOME") {
        return Ok(PathBuf::from(path));
    }
-    if let Some(path) = std::env::var_os("HOME") {
-        return Ok(PathBuf::from(path).join(".claw"));
-    }
-    if cfg!(target_os = "windows") {
-        if let Some(path) = std::env::var_os("USERPROFILE") {
-            return Ok(PathBuf::from(path).join(".claw"));
-        }
-    }
-    Err(io::Error::new(io::ErrorKind::NotFound, "HOME or USERPROFILE is not set"))
+    let home = std::env::var_os("HOME")
+        .or_else(|| std::env::var_os("USERPROFILE"))
+        .ok_or_else(|| {
+            io::Error::new(
+                io::ErrorKind::NotFound,
+                "HOME is not set (on Windows, set USERPROFILE or HOME, \
+                 or use CLAW_CONFIG_HOME to point directly at the config directory)",
+            )
+        })?;
+    Ok(PathBuf::from(home).join(".claw"))
 }

 fn read_credentials_root(path: &PathBuf) -> io::Result<Map<String, Value>> {
@@ -448,7 +456,7 @@ fn decode_hex(byte: u8) -> Result<u8, String> {
        b'0'..=b'9' => Ok(byte - b'0'),
        b'a'..=b'f' => Ok(byte - b'a' + 10),
        b'A'..=b'F' => Ok(byte - b'A' + 10),
-        _ => Err(format!("invalid percent-encoding byte: {byte}")),
+        _ => Err(format!("invalid percent byte: {byte}")),
    }
 }

@@ -0,0 +1,551 @@
+#![allow(
+    clippy::match_wildcard_for_single_variants,
+    clippy::must_use_candidate,
+    clippy::uninlined_format_args
+)]
+//! Permission enforcement layer that gates tool execution based on the
+//! active `PermissionPolicy`.
+
+use crate::permissions::{PermissionMode, PermissionOutcome, PermissionPolicy};
+use serde::{Deserialize, Serialize};
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(tag = "outcome")]
+pub enum EnforcementResult {
+    /// Tool execution is allowed.
+    Allowed,
+    /// Tool execution was denied due to insufficient permissions.
+    Denied {
+        tool: String,
+        active_mode: String,
+        required_mode: String,
+        reason: String,
+    },
+}
+
+#[derive(Debug, Clone, PartialEq)]
+pub struct PermissionEnforcer {
+    policy: PermissionPolicy,
+}
+
+impl PermissionEnforcer {
+    #[must_use]
+    pub fn new(policy: PermissionPolicy) -> Self {
+        Self { policy }
+    }
+
+    /// Check whether a tool can be executed under the current permission policy.
+    /// Auto-denies when prompting is required but no prompter is provided.
+    pub fn check(&self, tool_name: &str, input: &str) -> EnforcementResult {
+        // When the active mode is Prompt, defer to the caller's interactive
+        // prompt flow rather than hard-denying (the enforcer has no prompter).
+        if self.policy.active_mode() == PermissionMode::Prompt {
+            return EnforcementResult::Allowed;
+        }
+
+        let outcome = self.policy.authorize(tool_name, input, None);
+
+        match outcome {
+            PermissionOutcome::Allow => EnforcementResult::Allowed,
+            PermissionOutcome::Deny { reason } => {
+                let active_mode = self.policy.active_mode();
+                let required_mode = self.policy.required_mode_for(tool_name);
+                EnforcementResult::Denied {
+                    tool: tool_name.to_owned(),
+                    active_mode: active_mode.as_str().to_owned(),
+                    required_mode: required_mode.as_str().to_owned(),
+                    reason,
+                }
+            }
+        }
+    }
+
+    #[must_use]
+    pub fn is_allowed(&self, tool_name: &str, input: &str) -> bool {
+        matches!(self.check(tool_name, input), EnforcementResult::Allowed)
+    }
+
+    #[must_use]
+    pub fn active_mode(&self) -> PermissionMode {
+        self.policy.active_mode()
+    }
+
+    /// Classify a file operation against workspace boundaries.
+    pub fn check_file_write(&self, path: &str, workspace_root: &str) -> EnforcementResult {
+        let mode = self.policy.active_mode();
+
+        match mode {
+            PermissionMode::ReadOnly => EnforcementResult::Denied {
+                tool: "write_file".to_owned(),
+                active_mode: mode.as_str().to_owned(),
+                required_mode: PermissionMode::WorkspaceWrite.as_str().to_owned(),
+                reason: format!("file writes are not allowed in '{}' mode", mode.as_str()),
+            },
+            PermissionMode::WorkspaceWrite => {
+                if is_within_workspace(path, workspace_root) {
+                    EnforcementResult::Allowed
+                } else {
+                    EnforcementResult::Denied {
+                        tool: "write_file".to_owned(),
+                        active_mode: mode.as_str().to_owned(),
+                        required_mode: PermissionMode::DangerFullAccess.as_str().to_owned(),
+                        reason: format!(
+                            "path '{}' is outside workspace root '{}'",
+                            path, workspace_root
+                        ),
+                    }
+                }
+            }
+            // Allow and DangerFullAccess permit all writes
+            PermissionMode::Allow | PermissionMode::DangerFullAccess => EnforcementResult::Allowed,
+            PermissionMode::Prompt => EnforcementResult::Denied {
+                tool: "write_file".to_owned(),
+                active_mode: mode.as_str().to_owned(),
+                required_mode: PermissionMode::WorkspaceWrite.as_str().to_owned(),
+                reason: "file write requires confirmation in prompt mode".to_owned(),
+            },
+        }
+    }
+
+    /// Check if a bash command should be allowed based on current mode.
+    pub fn check_bash(&self, command: &str) -> EnforcementResult {
+        let mode = self.policy.active_mode();
+
+        match mode {
+            PermissionMode::ReadOnly => {
+                if is_read_only_command(command) {
+                    EnforcementResult::Allowed
+                } else {
+                    EnforcementResult::Denied {
+                        tool: "bash".to_owned(),
+                        active_mode: mode.as_str().to_owned(),
+                        required_mode: PermissionMode::WorkspaceWrite.as_str().to_owned(),
+                        reason: format!(
+                            "command may modify state; not allowed in '{}' mode",
+                            mode.as_str()
+                        ),
+                    }
+                }
+            }
+            PermissionMode::Prompt => EnforcementResult::Denied {
+                tool: "bash".to_owned(),
+                active_mode: mode.as_str().to_owned(),
+                required_mode: PermissionMode::DangerFullAccess.as_str().to_owned(),
+                reason: "bash requires confirmation in prompt mode".to_owned(),
+            },
+            // WorkspaceWrite, Allow, DangerFullAccess: permit bash
+            _ => EnforcementResult::Allowed,
+        }
+    }
+}
+
+/// Simple workspace boundary check via string prefix.
+fn is_within_workspace(path: &str, workspace_root: &str) -> bool {
+    let normalized = if path.starts_with('/') {
+        path.to_owned()
+    } else {
+        format!("{workspace_root}/{path}")
+    };
+
+    let root = if workspace_root.ends_with('/') {
+        workspace_root.to_owned()
+    } else {
+        format!("{workspace_root}/")
+    };
+
+    normalized.starts_with(&root) || normalized == workspace_root.trim_end_matches('/')
+}
+
+/// Conservative heuristic: is this bash command read-only?
+fn is_read_only_command(command: &str) -> bool {
+    let first_token = command
+        .split_whitespace()
+        .next()
+        .unwrap_or("")
+        .rsplit('/')
+        .next()
+        .unwrap_or("");
+
+    matches!(
+        first_token,
+        "cat"
+            | "head"
+            | "tail"
+            | "less"
+            | "more"
+            | "wc"
+            | "ls"
+            | "find"
+            | "grep"
+            | "rg"
+            | "awk"
+            | "sed"
+            | "echo"
+            | "printf"
+            | "which"
+            | "where"
+            | "whoami"
+            | "pwd"
+            | "env"
+            | "printenv"
+            | "date"
+            | "cal"
+            | "df"
+            | "du"
+            | "free"
+            | "uptime"
+            | "uname"
+            | "file"
+            | "stat"
+            | "diff"
+            | "sort"
+            | "uniq"
+            | "tr"
+            | "cut"
+            | "paste"
+            | "tee"
+            | "xargs"
+            | "test"
+            | "true"
+            | "false"
+            | "type"
+            | "readlink"
+            | "realpath"
+            | "basename"
+            | "dirname"
+            | "sha256sum"
+            | "md5sum"
+            | "b3sum"
+            | "xxd"
+            | "hexdump"
+            | "od"
+            | "strings"
+            | "tree"
+            | "jq"
+            | "yq"
+            | "python3"
+            | "python"
+            | "node"
+            | "ruby"
+            | "cargo"
+            | "rustc"
+            | "git"
+            | "gh"
+    ) && !command.contains("-i ")
+        && !command.contains("--in-place")
+        && !command.contains(" > ")
+        && !command.contains(" >> ")
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    fn make_enforcer(mode: PermissionMode) -> PermissionEnforcer {
+        let policy = PermissionPolicy::new(mode);
+        PermissionEnforcer::new(policy)
+    }
+
+    #[test]
+    fn allow_mode_permits_everything() {
+        let enforcer = make_enforcer(PermissionMode::Allow);
+        assert!(enforcer.is_allowed("bash", ""));
+        assert!(enforcer.is_allowed("write_file", ""));
+        assert!(enforcer.is_allowed("edit_file", ""));
+        assert_eq!(
+            enforcer.check_file_write("/outside/path", "/workspace"),
+            EnforcementResult::Allowed
+        );
+        assert_eq!(enforcer.check_bash("rm -rf /"), EnforcementResult::Allowed);
+    }
+
+    #[test]
+    fn read_only_denies_writes() {
+        let policy = PermissionPolicy::new(PermissionMode::ReadOnly)
+            .with_tool_requirement("read_file", PermissionMode::ReadOnly)
+            .with_tool_requirement("grep_search", PermissionMode::ReadOnly)
+            .with_tool_requirement("write_file", PermissionMode::WorkspaceWrite);
+
+        let enforcer = PermissionEnforcer::new(policy);
+        assert!(enforcer.is_allowed("read_file", ""));
+        assert!(enforcer.is_allowed("grep_search", ""));
+
+        // write_file requires WorkspaceWrite but we're in ReadOnly
+        let result = enforcer.check("write_file", "");
+        assert!(matches!(result, EnforcementResult::Denied { .. }));
+
+        let result = enforcer.check_file_write("/workspace/file.rs", "/workspace");
+        assert!(matches!(result, EnforcementResult::Denied { .. }));
+    }
+
+    #[test]
+    fn read_only_allows_read_commands() {
+        let enforcer = make_enforcer(PermissionMode::ReadOnly);
+        assert_eq!(
+            enforcer.check_bash("cat src/main.rs"),
+            EnforcementResult::Allowed
+        );
+        assert_eq!(
+            enforcer.check_bash("grep -r 'pattern' ."),
+            EnforcementResult::Allowed
+        );
+        assert_eq!(enforcer.check_bash("ls -la"), EnforcementResult::Allowed);
+    }
+
+    #[test]
+    fn read_only_denies_write_commands() {
+        let enforcer = make_enforcer(PermissionMode::ReadOnly);
+        let result = enforcer.check_bash("rm file.txt");
+        assert!(matches!(result, EnforcementResult::Denied { .. }));
+    }
+
+    #[test]
+    fn workspace_write_allows_within_workspace() {
+        let enforcer = make_enforcer(PermissionMode::WorkspaceWrite);
+        let result = enforcer.check_file_write("/workspace/src/main.rs", "/workspace");
+        assert_eq!(result, EnforcementResult::Allowed);
+    }
+
+    #[test]
+    fn workspace_write_denies_outside_workspace() {
+        let enforcer = make_enforcer(PermissionMode::WorkspaceWrite);
+        let result = enforcer.check_file_write("/etc/passwd", "/workspace");
+        assert!(matches!(result, EnforcementResult::Denied { .. }));
+    }
+
+    #[test]
+    fn prompt_mode_denies_without_prompter() {
+        let enforcer = make_enforcer(PermissionMode::Prompt);
+        let result = enforcer.check_bash("echo test");
+        assert!(matches!(result, EnforcementResult::Denied { .. }));
+
+        let result = enforcer.check_file_write("/workspace/file.rs", "/workspace");
+        assert!(matches!(result, EnforcementResult::Denied { .. }));
+    }
+
+    #[test]
+    fn workspace_boundary_check() {
+        assert!(is_within_workspace("/workspace/src/main.rs", "/workspace"));
+        assert!(is_within_workspace("/workspace", "/workspace"));
+        assert!(!is_within_workspace("/etc/passwd", "/workspace"));
+        assert!(!is_within_workspace("/workspacex/hack", "/workspace"));
+    }
+
+    #[test]
+    fn read_only_command_heuristic() {
+        assert!(is_read_only_command("cat file.txt"));
+        assert!(is_read_only_command("grep pattern file"));
+        assert!(is_read_only_command("git log --oneline"));
+        assert!(!is_read_only_command("rm file.txt"));
+        assert!(!is_read_only_command("echo test > file.txt"));
+        assert!(!is_read_only_command("sed -i 's/a/b/' file"));
+    }
+
+    #[test]
+    fn active_mode_returns_policy_mode() {
+        // given
+        let modes = [
+            PermissionMode::ReadOnly,
+            PermissionMode::WorkspaceWrite,
+            PermissionMode::DangerFullAccess,
+            PermissionMode::Prompt,
+            PermissionMode::Allow,
+        ];
+
+        // when
+        let active_modes: Vec<_> = modes
+            .into_iter()
+            .map(|mode| make_enforcer(mode).active_mode())
+            .collect();
+
+        // then
+        assert_eq!(active_modes, modes);
+    }
+
+    #[test]
+    fn danger_full_access_permits_file_writes_and_bash() {
+        // given
+        let enforcer = make_enforcer(PermissionMode::DangerFullAccess);
+
+        // when
+        let file_result = enforcer.check_file_write("/outside/workspace/file.txt", "/workspace");
+        let bash_result = enforcer.check_bash("rm -rf /tmp/scratch");
+
+        // then
+        assert_eq!(file_result, EnforcementResult::Allowed);
+        assert_eq!(bash_result, EnforcementResult::Allowed);
+    }
+
+    #[test]
+    fn check_denied_payload_contains_tool_and_modes() {
+        // given
+        let policy = PermissionPolicy::new(PermissionMode::ReadOnly)
+            .with_tool_requirement("write_file", PermissionMode::WorkspaceWrite);
+        let enforcer = PermissionEnforcer::new(policy);
+
+        // when
+        let result = enforcer.check("write_file", "{}");
+
+        // then
+        match result {
+            EnforcementResult::Denied {
+                tool,
+                active_mode,
+                required_mode,
+                reason,
+            } => {
+                assert_eq!(tool, "write_file");
+                assert_eq!(active_mode, "read-only");
+                assert_eq!(required_mode, "workspace-write");
+                assert!(reason.contains("requires workspace-write permission"));
+            }
+            other => panic!("expected denied result, got {other:?}"),
+        }
+    }
+
+    #[test]
+    fn workspace_write_relative_path_resolved() {
+        // given
+        let enforcer = make_enforcer(PermissionMode::WorkspaceWrite);
+
+        // when
+        let result = enforcer.check_file_write("src/main.rs", "/workspace");
+
+        // then
+        assert_eq!(result, EnforcementResult::Allowed);
+    }
+
+    #[test]
+    fn workspace_root_with_trailing_slash() {
+        // given
+        let enforcer = make_enforcer(PermissionMode::WorkspaceWrite);
+
+        // when
+        let result = enforcer.check_file_write("/workspace/src/main.rs", "/workspace/");
+
+        // then
+        assert_eq!(result, EnforcementResult::Allowed);
+    }
+
+    #[test]
+    fn workspace_root_equality() {
+        // given
+        let root = "/workspace/";
+
+        // when
+        let equal_to_root = is_within_workspace("/workspace", root);
+
+        // then
+        assert!(equal_to_root);
+    }
+
+    #[test]
+    fn bash_heuristic_full_path_prefix() {
+        // given
+        let full_path_command = "/usr/bin/cat Cargo.toml";
+        let git_path_command = "/usr/local/bin/git status";
+
+        // when
+        let cat_result = is_read_only_command(full_path_command);
+        let git_result = is_read_only_command(git_path_command);
+
+        // then
+        assert!(cat_result);
+        assert!(git_result);
+    }
+
+    #[test]
+    fn bash_heuristic_redirects_block_read_only_commands() {
+        // given
+        let overwrite = "cat Cargo.toml > out.txt";
+        let append = "echo test >> out.txt";
+
+        // when
+        let overwrite_result = is_read_only_command(overwrite);
+        let append_result = is_read_only_command(append);
+
+        // then
+        assert!(!overwrite_result);
+        assert!(!append_result);
+    }
+
+    #[test]
+    fn bash_heuristic_in_place_flag_blocks() {
+        // given
+        let interactive_python = "python -i script.py";
+        let in_place_sed = "sed --in-place 's/a/b/' file.txt";
+
+        // when
+        let interactive_result = is_read_only_command(interactive_python);
+        let in_place_result = is_read_only_command(in_place_sed);
+
+        // then
+        assert!(!interactive_result);
+        assert!(!in_place_result);
+    }
+
+    #[test]
+    fn bash_heuristic_empty_command() {
+        // given
+        let empty = "";
+        let whitespace = "   ";
+
+        // when
+        let empty_result = is_read_only_command(empty);
+        let whitespace_result = is_read_only_command(whitespace);
+
+        // then
+        assert!(!empty_result);
+        assert!(!whitespace_result);
+    }
+
+    #[test]
+    fn prompt_mode_check_bash_denied_payload_fields() {
+        // given
+        let enforcer = make_enforcer(PermissionMode::Prompt);
+
+        // when
+        let result = enforcer.check_bash("git status");
+
+        // then
+        match result {
+            EnforcementResult::Denied {
+                tool,
+                active_mode,
+                required_mode,
+                reason,
+            } => {
+                assert_eq!(tool, "bash");
+                assert_eq!(active_mode, "prompt");
+                assert_eq!(required_mode, "danger-full-access");
+                assert_eq!(reason, "bash requires confirmation in prompt mode");
+            }
+            other => panic!("expected denied result, got {other:?}"),
+        }
+    }
+
+    #[test]
+    fn read_only_check_file_write_denied_payload() {
+        // given
+        let enforcer = make_enforcer(PermissionMode::ReadOnly);
+
+        // when
+        let result = enforcer.check_file_write("/workspace/file.txt", "/workspace");
+
+        // then
+        match result {
+            EnforcementResult::Denied {
+                tool,
+                active_mode,
+                required_mode,
+                reason,
+            } => {
+                assert_eq!(tool, "write_file");
+                assert_eq!(active_mode, "read-only");
+                assert_eq!(required_mode, "workspace-write");
+                assert!(reason.contains("file writes are not allowed"));
+            }
+            other => panic!("expected denied result, got {other:?}"),
+        }
+    }
+}
@@ -1,5 +1,10 @@
 use std::collections::BTreeMap;

+use serde_json::Value;
+
+use crate::config::RuntimePermissionRuleConfig;
+
+/// Permission level assigned to a tool invocation or runtime session.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
 pub enum PermissionMode {
    ReadOnly,
@@ -22,34 +27,81 @@ impl PermissionMode {
    }
 }

+/// Hook-provided override applied before standard permission evaluation.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum PermissionOverride {
+    Allow,
+    Deny,
+    Ask,
+}
+
+/// Additional permission context supplied by hooks or higher-level orchestration.
+#[derive(Debug, Clone, PartialEq, Eq, Default)]
+pub struct PermissionContext {
+    override_decision: Option<PermissionOverride>,
+    override_reason: Option<String>,
+}
+
+impl PermissionContext {
+    #[must_use]
+    pub fn new(
+        override_decision: Option<PermissionOverride>,
+        override_reason: Option<String>,
+    ) -> Self {
+        Self {
+            override_decision,
+            override_reason,
+        }
+    }
+
+    #[must_use]
+    pub fn override_decision(&self) -> Option<PermissionOverride> {
+        self.override_decision
+    }
+
+    #[must_use]
+    pub fn override_reason(&self) -> Option<&str> {
+        self.override_reason.as_deref()
+    }
+}
+
+/// Full authorization request presented to a permission prompt.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct PermissionRequest {
    pub tool_name: String,
    pub input: String,
    pub current_mode: PermissionMode,
    pub required_mode: PermissionMode,
+    pub reason: Option<String>,
 }

+/// User-facing decision returned by a [`PermissionPrompter`].
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub enum PermissionPromptDecision {
    Allow,
    Deny { reason: String },
 }

+/// Prompting interface used when policy requires interactive approval.
 pub trait PermissionPrompter {
    fn decide(&mut self, request: &PermissionRequest) -> PermissionPromptDecision;
 }

+/// Final authorization result after evaluating static rules and prompts.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub enum PermissionOutcome {
    Allow,
    Deny { reason: String },
 }

+/// Evaluates permission mode requirements plus allow/deny/ask rules.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct PermissionPolicy {
    active_mode: PermissionMode,
    tool_requirements: BTreeMap<String, PermissionMode>,
+    allow_rules: Vec<PermissionRule>,
+    deny_rules: Vec<PermissionRule>,
+    ask_rules: Vec<PermissionRule>,
 }

 impl PermissionPolicy {
@@ -58,6 +110,9 @@ impl PermissionPolicy {
        Self {
            active_mode,
            tool_requirements: BTreeMap::new(),
+            allow_rules: Vec::new(),
+            deny_rules: Vec::new(),
+            ask_rules: Vec::new(),
        }
    }

@@ -72,6 +127,26 @@ impl PermissionPolicy {
        self
    }

+    #[must_use]
+    pub fn with_permission_rules(mut self, config: &RuntimePermissionRuleConfig) -> Self {
+        self.allow_rules = config
+            .allow()
+            .iter()
+            .map(|rule| PermissionRule::parse(rule))
+            .collect();
+        self.deny_rules = config
+            .deny()
+            .iter()
+            .map(|rule| PermissionRule::parse(rule))
+            .collect();
+        self.ask_rules = config
+            .ask()
+            .iter()
+            .map(|rule| PermissionRule::parse(rule))
+            .collect();
+        self
+    }
+
    #[must_use]
    pub fn active_mode(&self) -> PermissionMode {
        self.active_mode
@@ -90,38 +165,121 @@ impl PermissionPolicy {
        &self,
        tool_name: &str,
        input: &str,
-        mut prompter: Option<&mut dyn PermissionPrompter>,
+        prompter: Option<&mut dyn PermissionPrompter>,
    ) -> PermissionOutcome {
-        let current_mode = self.active_mode();
-        let required_mode = self.required_mode_for(tool_name);
-        if current_mode == PermissionMode::Allow || current_mode >= required_mode {
-            return PermissionOutcome::Allow;
+        self.authorize_with_context(tool_name, input, &PermissionContext::default(), prompter)
+    }
+
+    #[must_use]
+    #[allow(clippy::too_many_lines)]
+    pub fn authorize_with_context(
+        &self,
+        tool_name: &str,
+        input: &str,
+        context: &PermissionContext,
+        prompter: Option<&mut dyn PermissionPrompter>,
+    ) -> PermissionOutcome {
+        if let Some(rule) = Self::find_matching_rule(&self.deny_rules, tool_name, input) {
+            return PermissionOutcome::Deny {
+                reason: format!(
+                    "Permission to use {tool_name} has been denied by rule '{}'",
+                    rule.raw
+                ),
+            };
        }

-        let request = PermissionRequest {
-            tool_name: tool_name.to_string(),
-            input: input.to_string(),
-            current_mode,
-            required_mode,
-        };
+        let current_mode = self.active_mode();
+        let required_mode = self.required_mode_for(tool_name);
+        let ask_rule = Self::find_matching_rule(&self.ask_rules, tool_name, input);
+        let allow_rule = Self::find_matching_rule(&self.allow_rules, tool_name, input);
+
+        match context.override_decision() {
+            Some(PermissionOverride::Deny) => {
+                return PermissionOutcome::Deny {
+                    reason: context.override_reason().map_or_else(
+                        || format!("tool '{tool_name}' denied by hook"),
+                        ToOwned::to_owned,
+                    ),
+                };
+            }
+            Some(PermissionOverride::Ask) => {
+                let reason = context.override_reason().map_or_else(
+                    || format!("tool '{tool_name}' requires approval due to hook guidance"),
+                    ToOwned::to_owned,
+                );
+                return Self::prompt_or_deny(
+                    tool_name,
+                    input,
+                    current_mode,
+                    required_mode,
+                    Some(reason),
+                    prompter,
+                );
+            }
+            Some(PermissionOverride::Allow) => {
+                if let Some(rule) = ask_rule {
+                    let reason = format!(
+                        "tool '{tool_name}' requires approval due to ask rule '{}'",
+                        rule.raw
+                    );
+                    return Self::prompt_or_deny(
+                        tool_name,
+                        input,
+                        current_mode,
+                        required_mode,
+                        Some(reason),
+                        prompter,
+                    );
+                }
+                if allow_rule.is_some()
+                    || current_mode == PermissionMode::Allow
+                    || current_mode >= required_mode
+                {
+                    return PermissionOutcome::Allow;
+                }
+            }
+            None => {}
+        }
+
+        if let Some(rule) = ask_rule {
+            let reason = format!(
+                "tool '{tool_name}' requires approval due to ask rule '{}'",
+                rule.raw
+            );
+            return Self::prompt_or_deny(
+                tool_name,
+                input,
+                current_mode,
+                required_mode,
+                Some(reason),
+                prompter,
+            );
+        }
+
+        if allow_rule.is_some()
+            || current_mode == PermissionMode::Allow
+            || current_mode >= required_mode
+        {
+            return PermissionOutcome::Allow;
+        }

        if current_mode == PermissionMode::Prompt
            || (current_mode == PermissionMode::WorkspaceWrite
                && required_mode == PermissionMode::DangerFullAccess)
        {
-            return match prompter.as_mut() {
-                Some(prompter) => match prompter.decide(&request) {
-                    PermissionPromptDecision::Allow => PermissionOutcome::Allow,
-                    PermissionPromptDecision::Deny { reason } => PermissionOutcome::Deny { reason },
-                },
-                None => PermissionOutcome::Deny {
-                    reason: format!(
-                        "tool '{tool_name}' requires approval to escalate from {} to {}",
-                        current_mode.as_str(),
-                        required_mode.as_str()
-                    ),
-                },
-            };
+            let reason = Some(format!(
+                "tool '{tool_name}' requires approval to escalate from {} to {}",
+                current_mode.as_str(),
+                required_mode.as_str()
+            ));
+            return Self::prompt_or_deny(
+                tool_name,
+                input,
+                current_mode,
+                required_mode,
+                reason,
+                prompter,
+            );
        }

        PermissionOutcome::Deny {
@@ -132,14 +290,191 @@ impl PermissionPolicy {
            ),
        }
    }
+
+    fn prompt_or_deny(
+        tool_name: &str,
+        input: &str,
+        current_mode: PermissionMode,
+        required_mode: PermissionMode,
+        reason: Option<String>,
+        mut prompter: Option<&mut dyn PermissionPrompter>,
+    ) -> PermissionOutcome {
+        let request = PermissionRequest {
+            tool_name: tool_name.to_string(),
+            input: input.to_string(),
+            current_mode,
+            required_mode,
+            reason: reason.clone(),
+        };
+
+        match prompter.as_mut() {
+            Some(prompter) => match prompter.decide(&request) {
+                PermissionPromptDecision::Allow => PermissionOutcome::Allow,
+                PermissionPromptDecision::Deny { reason } => PermissionOutcome::Deny { reason },
+            },
+            None => PermissionOutcome::Deny {
+                reason: reason.unwrap_or_else(|| {
+                    format!(
+                        "tool '{tool_name}' requires approval to run while mode is {}",
+                        current_mode.as_str()
+                    )
+                }),
+            },
+        }
+    }
+
+    fn find_matching_rule<'a>(
+        rules: &'a [PermissionRule],
+        tool_name: &str,
+        input: &str,
+    ) -> Option<&'a PermissionRule> {
+        rules.iter().find(|rule| rule.matches(tool_name, input))
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+struct PermissionRule {
+    raw: String,
+    tool_name: String,
+    matcher: PermissionRuleMatcher,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+enum PermissionRuleMatcher {
+    Any,
+    Exact(String),
+    Prefix(String),
+}
+
+impl PermissionRule {
+    fn parse(raw: &str) -> Self {
+        let trimmed = raw.trim();
+        let open = find_first_unescaped(trimmed, '(');
+        let close = find_last_unescaped(trimmed, ')');
+
+        if let (Some(open), Some(close)) = (open, close) {
+            if close == trimmed.len() - 1 && open < close {
+                let tool_name = trimmed[..open].trim();
+                let content = &trimmed[open + 1..close];
+                if !tool_name.is_empty() {
+                    let matcher = parse_rule_matcher(content);
+                    return Self {
+                        raw: trimmed.to_string(),
+                        tool_name: tool_name.to_string(),
+                        matcher,
+                    };
+                }
+            }
+        }
+
+        Self {
+            raw: trimmed.to_string(),
+            tool_name: trimmed.to_string(),
+            matcher: PermissionRuleMatcher::Any,
+        }
+    }
+
+    fn matches(&self, tool_name: &str, input: &str) -> bool {
+        if self.tool_name != tool_name {
+            return false;
+        }
+
+        match &self.matcher {
+            PermissionRuleMatcher::Any => true,
+            PermissionRuleMatcher::Exact(expected) => {
+                extract_permission_subject(input).is_some_and(|candidate| candidate == *expected)
+            }
+            PermissionRuleMatcher::Prefix(prefix) => extract_permission_subject(input)
+                .is_some_and(|candidate| candidate.starts_with(prefix)),
+        }
+    }
+}
+
+fn parse_rule_matcher(content: &str) -> PermissionRuleMatcher {
+    let unescaped = unescape_rule_content(content.trim());
+    if unescaped.is_empty() || unescaped == "*" {
+        PermissionRuleMatcher::Any
+    } else if let Some(prefix) = unescaped.strip_suffix(":*") {
+        PermissionRuleMatcher::Prefix(prefix.to_string())
+    } else {
+        PermissionRuleMatcher::Exact(unescaped)
+    }
+}
+
+fn unescape_rule_content(content: &str) -> String {
+    content
+        .replace(r"\(", "(")
+        .replace(r"\)", ")")
+        .replace(r"\\", r"\")
+}
+
+fn find_first_unescaped(value: &str, needle: char) -> Option<usize> {
+    let mut escaped = false;
+    for (idx, ch) in value.char_indices() {
+        if ch == '\\' {
+            escaped = !escaped;
+            continue;
+        }
+        if ch == needle && !escaped {
+            return Some(idx);
+        }
+        escaped = false;
+    }
+    None
+}
+
+fn find_last_unescaped(value: &str, needle: char) -> Option<usize> {
+    let chars = value.char_indices().collect::<Vec<_>>();
+    for (pos, (idx, ch)) in chars.iter().enumerate().rev() {
+        if *ch != needle {
+            continue;
+        }
+        let mut backslashes = 0;
+        for (_, prev) in chars[..pos].iter().rev() {
+            if *prev == '\\' {
+                backslashes += 1;
+            } else {
+                break;
+            }
+        }
+        if backslashes % 2 == 0 {
+            return Some(*idx);
+        }
+    }
+    None
+}
+
+fn extract_permission_subject(input: &str) -> Option<String> {
+    let parsed = serde_json::from_str::<Value>(input).ok();
+    if let Some(Value::Object(object)) = parsed {
+        for key in [
+            "command",
+            "path",
+            "file_path",
+            "filePath",
+            "notebook_path",
+            "notebookPath",
+            "url",
+            "pattern",
+            "code",
+            "message",
+        ] {
+            if let Some(value) = object.get(key).and_then(Value::as_str) {
+                return Some(value.to_string());
+            }
+        }
+    }
+
+    (!input.trim().is_empty()).then(|| input.to_string())
 }

 #[cfg(test)]
 mod tests {
    use super::{
-        PermissionMode, PermissionOutcome, PermissionPolicy, PermissionPromptDecision,
-        PermissionPrompter, PermissionRequest,
+        PermissionContext, PermissionMode, PermissionOutcome, PermissionOverride, PermissionPolicy,
+        PermissionPromptDecision, PermissionPrompter, PermissionRequest,
    };
+    use crate::config::RuntimePermissionRuleConfig;

    struct RecordingPrompter {
        seen: Vec<PermissionRequest>,
@@ -229,4 +564,120 @@ mod tests {
            PermissionOutcome::Deny { reason } if reason == "not now"
        ));
    }
+
+    #[test]
+    fn applies_rule_based_denials_and_allows() {
+        let rules = RuntimePermissionRuleConfig::new(
+            vec!["bash(git:*)".to_string()],
+            vec!["bash(rm -rf:*)".to_string()],
+            Vec::new(),
+        );
+        let policy = PermissionPolicy::new(PermissionMode::ReadOnly)
+            .with_tool_requirement("bash", PermissionMode::DangerFullAccess)
+            .with_permission_rules(&rules);
+
+        assert_eq!(
+            policy.authorize("bash", r#"{"command":"git status"}"#, None),
+            PermissionOutcome::Allow
+        );
+        assert!(matches!(
+            policy.authorize("bash", r#"{"command":"rm -rf /tmp/x"}"#, None),
+            PermissionOutcome::Deny { reason } if reason.contains("denied by rule")
+        ));
+    }
+
+    #[test]
+    fn ask_rules_force_prompt_even_when_mode_allows() {
+        let rules = RuntimePermissionRuleConfig::new(
+            Vec::new(),
+            Vec::new(),
+            vec!["bash(git:*)".to_string()],
+        );
+        let policy = PermissionPolicy::new(PermissionMode::DangerFullAccess)
+            .with_tool_requirement("bash", PermissionMode::DangerFullAccess)
+            .with_permission_rules(&rules);
+        let mut prompter = RecordingPrompter {
+            seen: Vec::new(),
+            allow: true,
+        };
+
+        let outcome = policy.authorize("bash", r#"{"command":"git status"}"#, Some(&mut prompter));
+
+        assert_eq!(outcome, PermissionOutcome::Allow);
+        assert_eq!(prompter.seen.len(), 1);
+        assert!(prompter.seen[0]
+            .reason
+            .as_deref()
+            .is_some_and(|reason| reason.contains("ask rule")));
+    }
+
+    #[test]
+    fn hook_allow_still_respects_ask_rules() {
+        let rules = RuntimePermissionRuleConfig::new(
+            Vec::new(),
+            Vec::new(),
+            vec!["bash(git:*)".to_string()],
+        );
+        let policy = PermissionPolicy::new(PermissionMode::ReadOnly)
+            .with_tool_requirement("bash", PermissionMode::DangerFullAccess)
+            .with_permission_rules(&rules);
+        let context = PermissionContext::new(
+            Some(PermissionOverride::Allow),
+            Some("hook approved".to_string()),
+        );
+        let mut prompter = RecordingPrompter {
+            seen: Vec::new(),
+            allow: true,
+        };
+
+        let outcome = policy.authorize_with_context(
+            "bash",
+            r#"{"command":"git status"}"#,
+            &context,
+            Some(&mut prompter),
+        );
+
+        assert_eq!(outcome, PermissionOutcome::Allow);
+        assert_eq!(prompter.seen.len(), 1);
+    }
+
+    #[test]
+    fn hook_deny_short_circuits_permission_flow() {
+        let policy = PermissionPolicy::new(PermissionMode::DangerFullAccess)
+            .with_tool_requirement("bash", PermissionMode::DangerFullAccess);
+        let context = PermissionContext::new(
+            Some(PermissionOverride::Deny),
+            Some("blocked by hook".to_string()),
+        );
+
+        assert_eq!(
+            policy.authorize_with_context("bash", "{}", &context, None),
+            PermissionOutcome::Deny {
+                reason: "blocked by hook".to_string(),
+            }
+        );
+    }
+
+    #[test]
+    fn hook_ask_forces_prompt() {
+        let policy = PermissionPolicy::new(PermissionMode::DangerFullAccess)
+            .with_tool_requirement("bash", PermissionMode::DangerFullAccess);
+        let context = PermissionContext::new(
+            Some(PermissionOverride::Ask),
+            Some("hook requested confirmation".to_string()),
+        );
+        let mut prompter = RecordingPrompter {
+            seen: Vec::new(),
+            allow: true,
+        };
+
+        let outcome = policy.authorize_with_context("bash", "{}", &context, Some(&mut prompter));
+
+        assert_eq!(outcome, PermissionOutcome::Allow);
+        assert_eq!(prompter.seen.len(), 1);
+        assert_eq!(
+            prompter.seen[0].reason.as_deref(),
+            Some("hook requested confirmation")
+        );
+    }
 }
@@ -0,0 +1,533 @@
+#![allow(clippy::redundant_closure_for_method_calls)]
+use std::time::{SystemTime, UNIX_EPOCH};
+
+use serde::{Deserialize, Serialize};
+
+use crate::config::RuntimePluginConfig;
+use crate::mcp_tool_bridge::{McpResourceInfo, McpToolInfo};
+
+fn now_secs() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .unwrap_or_default()
+        .as_secs()
+}
+
+pub type ToolInfo = McpToolInfo;
+pub type ResourceInfo = McpResourceInfo;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum ServerStatus {
+    Healthy,
+    Degraded,
+    Failed,
+}
+
+impl std::fmt::Display for ServerStatus {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Healthy => write!(f, "healthy"),
+            Self::Degraded => write!(f, "degraded"),
+            Self::Failed => write!(f, "failed"),
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct ServerHealth {
+    pub server_name: String,
+    pub status: ServerStatus,
+    pub capabilities: Vec<String>,
+    pub last_error: Option<String>,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case", tag = "state")]
+pub enum PluginState {
+    Unconfigured,
+    Validated,
+    Starting,
+    Healthy,
+    Degraded {
+        healthy_servers: Vec<String>,
+        failed_servers: Vec<ServerHealth>,
+    },
+    Failed {
+        reason: String,
+    },
+    ShuttingDown,
+    Stopped,
+}
+
+impl PluginState {
+    #[must_use]
+    pub fn from_servers(servers: &[ServerHealth]) -> Self {
+        if servers.is_empty() {
+            return Self::Failed {
+                reason: "no servers available".to_string(),
+            };
+        }
+
+        let healthy_servers = servers
+            .iter()
+            .filter(|server| server.status != ServerStatus::Failed)
+            .map(|server| server.server_name.clone())
+            .collect::<Vec<_>>();
+        let failed_servers = servers
+            .iter()
+            .filter(|server| server.status == ServerStatus::Failed)
+            .cloned()
+            .collect::<Vec<_>>();
+        let has_degraded_server = servers
+            .iter()
+            .any(|server| server.status == ServerStatus::Degraded);
+
+        if failed_servers.is_empty() && !has_degraded_server {
+            Self::Healthy
+        } else if healthy_servers.is_empty() {
+            Self::Failed {
+                reason: format!("all {} servers failed", failed_servers.len()),
+            }
+        } else {
+            Self::Degraded {
+                healthy_servers,
+                failed_servers,
+            }
+        }
+    }
+}
+
+impl std::fmt::Display for PluginState {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Unconfigured => write!(f, "unconfigured"),
+            Self::Validated => write!(f, "validated"),
+            Self::Starting => write!(f, "starting"),
+            Self::Healthy => write!(f, "healthy"),
+            Self::Degraded { .. } => write!(f, "degraded"),
+            Self::Failed { .. } => write!(f, "failed"),
+            Self::ShuttingDown => write!(f, "shutting_down"),
+            Self::Stopped => write!(f, "stopped"),
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct PluginHealthcheck {
+    pub plugin_name: String,
+    pub state: PluginState,
+    pub servers: Vec<ServerHealth>,
+    pub last_check: u64,
+}
+
+impl PluginHealthcheck {
+    #[must_use]
+    pub fn new(plugin_name: impl Into<String>, servers: Vec<ServerHealth>) -> Self {
+        let state = PluginState::from_servers(&servers);
+        Self {
+            plugin_name: plugin_name.into(),
+            state,
+            servers,
+            last_check: now_secs(),
+        }
+    }
+
+    #[must_use]
+    pub fn degraded_mode(&self, discovery: &DiscoveryResult) -> Option<DegradedMode> {
+        match &self.state {
+            PluginState::Degraded {
+                healthy_servers,
+                failed_servers,
+            } => Some(DegradedMode {
+                available_tools: discovery
+                    .tools
+                    .iter()
+                    .map(|tool| tool.name.clone())
+                    .collect(),
+                unavailable_tools: failed_servers
+                    .iter()
+                    .flat_map(|server| server.capabilities.iter().cloned())
+                    .collect(),
+                reason: format!(
+                    "{} servers healthy, {} servers failed",
+                    healthy_servers.len(),
+                    failed_servers.len()
+                ),
+            }),
+            _ => None,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct DiscoveryResult {
+    pub tools: Vec<ToolInfo>,
+    pub resources: Vec<ResourceInfo>,
+    pub partial: bool,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct DegradedMode {
+    pub available_tools: Vec<String>,
+    pub unavailable_tools: Vec<String>,
+    pub reason: String,
+}
+
+impl DegradedMode {
+    #[must_use]
+    pub fn new(
+        available_tools: Vec<String>,
+        unavailable_tools: Vec<String>,
+        reason: impl Into<String>,
+    ) -> Self {
+        Self {
+            available_tools,
+            unavailable_tools,
+            reason: reason.into(),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum PluginLifecycleEvent {
+    ConfigValidated,
+    StartupHealthy,
+    StartupDegraded,
+    StartupFailed,
+    Shutdown,
+}
+
+impl std::fmt::Display for PluginLifecycleEvent {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::ConfigValidated => write!(f, "config_validated"),
+            Self::StartupHealthy => write!(f, "startup_healthy"),
+            Self::StartupDegraded => write!(f, "startup_degraded"),
+            Self::StartupFailed => write!(f, "startup_failed"),
+            Self::Shutdown => write!(f, "shutdown"),
+        }
+    }
+}
+
+pub trait PluginLifecycle {
+    fn validate_config(&self, config: &RuntimePluginConfig) -> Result<(), String>;
+    fn healthcheck(&self) -> PluginHealthcheck;
+    fn discover(&self) -> DiscoveryResult;
+    fn shutdown(&mut self) -> Result<(), String>;
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[derive(Debug, Clone)]
+    struct MockPluginLifecycle {
+        plugin_name: String,
+        valid_config: bool,
+        healthcheck: PluginHealthcheck,
+        discovery: DiscoveryResult,
+        shutdown_error: Option<String>,
+        shutdown_called: bool,
+    }
+
+    impl MockPluginLifecycle {
+        fn new(
+            plugin_name: &str,
+            valid_config: bool,
+            servers: Vec<ServerHealth>,
+            discovery: DiscoveryResult,
+            shutdown_error: Option<String>,
+        ) -> Self {
+            Self {
+                plugin_name: plugin_name.to_string(),
+                valid_config,
+                healthcheck: PluginHealthcheck::new(plugin_name, servers),
+                discovery,
+                shutdown_error,
+                shutdown_called: false,
+            }
+        }
+    }
+
+    impl PluginLifecycle for MockPluginLifecycle {
+        fn validate_config(&self, _config: &RuntimePluginConfig) -> Result<(), String> {
+            if self.valid_config {
+                Ok(())
+            } else {
+                Err(format!(
+                    "plugin `{}` failed configuration validation",
+                    self.plugin_name
+                ))
+            }
+        }
+
+        fn healthcheck(&self) -> PluginHealthcheck {
+            if self.shutdown_called {
+                PluginHealthcheck {
+                    plugin_name: self.plugin_name.clone(),
+                    state: PluginState::Stopped,
+                    servers: self.healthcheck.servers.clone(),
+                    last_check: now_secs(),
+                }
+            } else {
+                self.healthcheck.clone()
+            }
+        }
+
+        fn discover(&self) -> DiscoveryResult {
+            self.discovery.clone()
+        }
+
+        fn shutdown(&mut self) -> Result<(), String> {
+            if let Some(error) = &self.shutdown_error {
+                return Err(error.clone());
+            }
+
+            self.shutdown_called = true;
+            Ok(())
+        }
+    }
+
+    fn healthy_server(name: &str, capabilities: &[&str]) -> ServerHealth {
+        ServerHealth {
+            server_name: name.to_string(),
+            status: ServerStatus::Healthy,
+            capabilities: capabilities
+                .iter()
+                .map(|capability| capability.to_string())
+                .collect(),
+            last_error: None,
+        }
+    }
+
+    fn failed_server(name: &str, capabilities: &[&str], error: &str) -> ServerHealth {
+        ServerHealth {
+            server_name: name.to_string(),
+            status: ServerStatus::Failed,
+            capabilities: capabilities
+                .iter()
+                .map(|capability| capability.to_string())
+                .collect(),
+            last_error: Some(error.to_string()),
+        }
+    }
+
+    fn degraded_server(name: &str, capabilities: &[&str], error: &str) -> ServerHealth {
+        ServerHealth {
+            server_name: name.to_string(),
+            status: ServerStatus::Degraded,
+            capabilities: capabilities
+                .iter()
+                .map(|capability| capability.to_string())
+                .collect(),
+            last_error: Some(error.to_string()),
+        }
+    }
+
+    fn tool(name: &str) -> ToolInfo {
+        ToolInfo {
+            name: name.to_string(),
+            description: Some(format!("{name} tool")),
+            input_schema: None,
+        }
+    }
+
+    fn resource(name: &str, uri: &str) -> ResourceInfo {
+        ResourceInfo {
+            uri: uri.to_string(),
+            name: name.to_string(),
+            description: Some(format!("{name} resource")),
+            mime_type: Some("application/json".to_string()),
+        }
+    }
+
+    #[test]
+    fn full_lifecycle_happy_path() {
+        // given
+        let mut lifecycle = MockPluginLifecycle::new(
+            "healthy-plugin",
+            true,
+            vec![
+                healthy_server("alpha", &["search", "read"]),
+                healthy_server("beta", &["write"]),
+            ],
+            DiscoveryResult {
+                tools: vec![tool("search"), tool("read"), tool("write")],
+                resources: vec![resource("docs", "file:///docs")],
+                partial: false,
+            },
+            None,
+        );
+        let config = RuntimePluginConfig::default();
+
+        // when
+        let validation = lifecycle.validate_config(&config);
+        let healthcheck = lifecycle.healthcheck();
+        let discovery = lifecycle.discover();
+        let shutdown = lifecycle.shutdown();
+        let post_shutdown = lifecycle.healthcheck();
+
+        // then
+        assert_eq!(validation, Ok(()));
+        assert_eq!(healthcheck.state, PluginState::Healthy);
+        assert_eq!(healthcheck.plugin_name, "healthy-plugin");
+        assert_eq!(discovery.tools.len(), 3);
+        assert_eq!(discovery.resources.len(), 1);
+        assert!(!discovery.partial);
+        assert_eq!(shutdown, Ok(()));
+        assert_eq!(post_shutdown.state, PluginState::Stopped);
+    }
+
+    #[test]
+    fn degraded_startup_when_one_of_three_servers_fails() {
+        // given
+        let lifecycle = MockPluginLifecycle::new(
+            "degraded-plugin",
+            true,
+            vec![
+                healthy_server("alpha", &["search"]),
+                failed_server("beta", &["write"], "connection refused"),
+                healthy_server("gamma", &["read"]),
+            ],
+            DiscoveryResult {
+                tools: vec![tool("search"), tool("read")],
+                resources: vec![resource("alpha-docs", "file:///alpha")],
+                partial: true,
+            },
+            None,
+        );
+
+        // when
+        let healthcheck = lifecycle.healthcheck();
+        let discovery = lifecycle.discover();
+        let degraded_mode = healthcheck
+            .degraded_mode(&discovery)
+            .expect("degraded startup should expose degraded mode");
+
+        // then
+        match healthcheck.state {
+            PluginState::Degraded {
+                healthy_servers,
+                failed_servers,
+            } => {
+                assert_eq!(
+                    healthy_servers,
+                    vec!["alpha".to_string(), "gamma".to_string()]
+                );
+                assert_eq!(failed_servers.len(), 1);
+                assert_eq!(failed_servers[0].server_name, "beta");
+                assert_eq!(
+                    failed_servers[0].last_error.as_deref(),
+                    Some("connection refused")
+                );
+            }
+            other => panic!("expected degraded state, got {other:?}"),
+        }
+        assert!(discovery.partial);
+        assert_eq!(
+            degraded_mode.available_tools,
+            vec!["search".to_string(), "read".to_string()]
+        );
+        assert_eq!(degraded_mode.unavailable_tools, vec!["write".to_string()]);
+        assert_eq!(degraded_mode.reason, "2 servers healthy, 1 servers failed");
+    }
+
+    #[test]
+    fn degraded_server_status_keeps_server_usable() {
+        // given
+        let lifecycle = MockPluginLifecycle::new(
+            "soft-degraded-plugin",
+            true,
+            vec![
+                healthy_server("alpha", &["search"]),
+                degraded_server("beta", &["write"], "high latency"),
+            ],
+            DiscoveryResult {
+                tools: vec![tool("search"), tool("write")],
+                resources: Vec::new(),
+                partial: true,
+            },
+            None,
+        );
+
+        // when
+        let healthcheck = lifecycle.healthcheck();
+
+        // then
+        match healthcheck.state {
+            PluginState::Degraded {
+                healthy_servers,
+                failed_servers,
+            } => {
+                assert_eq!(
+                    healthy_servers,
+                    vec!["alpha".to_string(), "beta".to_string()]
+                );
+                assert!(failed_servers.is_empty());
+            }
+            other => panic!("expected degraded state, got {other:?}"),
+        }
+    }
+
+    #[test]
+    fn complete_failure_when_all_servers_fail() {
+        // given
+        let lifecycle = MockPluginLifecycle::new(
+            "failed-plugin",
+            true,
+            vec![
+                failed_server("alpha", &["search"], "timeout"),
+                failed_server("beta", &["read"], "handshake failed"),
+            ],
+            DiscoveryResult {
+                tools: Vec::new(),
+                resources: Vec::new(),
+                partial: false,
+            },
+            None,
+        );
+
+        // when
+        let healthcheck = lifecycle.healthcheck();
+        let discovery = lifecycle.discover();
+
+        // then
+        match &healthcheck.state {
+            PluginState::Failed { reason } => {
+                assert_eq!(reason, "all 2 servers failed");
+            }
+            other => panic!("expected failed state, got {other:?}"),
+        }
+        assert!(!discovery.partial);
+        assert!(discovery.tools.is_empty());
+        assert!(discovery.resources.is_empty());
+        assert!(healthcheck.degraded_mode(&discovery).is_none());
+    }
+
+    #[test]
+    fn graceful_shutdown() {
+        // given
+        let mut lifecycle = MockPluginLifecycle::new(
+            "shutdown-plugin",
+            true,
+            vec![healthy_server("alpha", &["search"])],
+            DiscoveryResult {
+                tools: vec![tool("search")],
+                resources: Vec::new(),
+                partial: false,
+            },
+            None,
+        );
+
+        // when
+        let shutdown = lifecycle.shutdown();
+        let post_shutdown = lifecycle.healthcheck();
+
+        // then
+        assert_eq!(shutdown, Ok(()));
+        assert_eq!(PluginLifecycleEvent::Shutdown.to_string(), "shutdown");
+        assert_eq!(post_shutdown.state, PluginState::Stopped);
+    }
+}
@@ -0,0 +1,581 @@
+use std::time::Duration;
+
+pub type GreenLevel = u8;
+
+const STALE_BRANCH_THRESHOLD: Duration = Duration::from_secs(60 * 60);
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct PolicyRule {
+    pub name: String,
+    pub condition: PolicyCondition,
+    pub action: PolicyAction,
+    pub priority: u32,
+}
+
+impl PolicyRule {
+    #[must_use]
+    pub fn new(
+        name: impl Into<String>,
+        condition: PolicyCondition,
+        action: PolicyAction,
+        priority: u32,
+    ) -> Self {
+        Self {
+            name: name.into(),
+            condition,
+            action,
+            priority,
+        }
+    }
+
+    #[must_use]
+    pub fn matches(&self, context: &LaneContext) -> bool {
+        self.condition.matches(context)
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum PolicyCondition {
+    And(Vec<PolicyCondition>),
+    Or(Vec<PolicyCondition>),
+    GreenAt { level: GreenLevel },
+    StaleBranch,
+    StartupBlocked,
+    LaneCompleted,
+    LaneReconciled,
+    ReviewPassed,
+    ScopedDiff,
+    TimedOut { duration: Duration },
+}
+
+impl PolicyCondition {
+    #[must_use]
+    pub fn matches(&self, context: &LaneContext) -> bool {
+        match self {
+            Self::And(conditions) => conditions
+                .iter()
+                .all(|condition| condition.matches(context)),
+            Self::Or(conditions) => conditions
+                .iter()
+                .any(|condition| condition.matches(context)),
+            Self::GreenAt { level } => context.green_level >= *level,
+            Self::StaleBranch => context.branch_freshness >= STALE_BRANCH_THRESHOLD,
+            Self::StartupBlocked => context.blocker == LaneBlocker::Startup,
+            Self::LaneCompleted => context.completed,
+            Self::LaneReconciled => context.reconciled,
+            Self::ReviewPassed => context.review_status == ReviewStatus::Approved,
+            Self::ScopedDiff => context.diff_scope == DiffScope::Scoped,
+            Self::TimedOut { duration } => context.branch_freshness >= *duration,
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum PolicyAction {
+    MergeToDev,
+    MergeForward,
+    RecoverOnce,
+    Escalate { reason: String },
+    CloseoutLane,
+    CleanupSession,
+    Reconcile { reason: ReconcileReason },
+    Notify { channel: String },
+    Block { reason: String },
+    Chain(Vec<PolicyAction>),
+}
+
+/// Why a lane was reconciled without further action.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum ReconcileReason {
+    /// Branch already merged into main — no PR needed.
+    AlreadyMerged,
+    /// Work superseded by another lane or direct commit.
+    Superseded,
+    /// PR would be empty — all changes already landed.
+    EmptyDiff,
+    /// Lane manually closed by operator.
+    ManualClose,
+}
+
+impl PolicyAction {
+    fn flatten_into(&self, actions: &mut Vec<PolicyAction>) {
+        match self {
+            Self::Chain(chained) => {
+                for action in chained {
+                    action.flatten_into(actions);
+                }
+            }
+            _ => actions.push(self.clone()),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum LaneBlocker {
+    None,
+    Startup,
+    External,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ReviewStatus {
+    Pending,
+    Approved,
+    Rejected,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum DiffScope {
+    Full,
+    Scoped,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct LaneContext {
+    pub lane_id: String,
+    pub green_level: GreenLevel,
+    pub branch_freshness: Duration,
+    pub blocker: LaneBlocker,
+    pub review_status: ReviewStatus,
+    pub diff_scope: DiffScope,
+    pub completed: bool,
+    pub reconciled: bool,
+}
+
+impl LaneContext {
+    #[must_use]
+    pub fn new(
+        lane_id: impl Into<String>,
+        green_level: GreenLevel,
+        branch_freshness: Duration,
+        blocker: LaneBlocker,
+        review_status: ReviewStatus,
+        diff_scope: DiffScope,
+        completed: bool,
+    ) -> Self {
+        Self {
+            lane_id: lane_id.into(),
+            green_level,
+            branch_freshness,
+            blocker,
+            review_status,
+            diff_scope,
+            completed,
+            reconciled: false,
+        }
+    }
+
+    /// Create a lane context that is already reconciled (no further action needed).
+    #[must_use]
+    pub fn reconciled(lane_id: impl Into<String>) -> Self {
+        Self {
+            lane_id: lane_id.into(),
+            green_level: 0,
+            branch_freshness: Duration::from_secs(0),
+            blocker: LaneBlocker::None,
+            review_status: ReviewStatus::Pending,
+            diff_scope: DiffScope::Full,
+            completed: true,
+            reconciled: true,
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct PolicyEngine {
+    rules: Vec<PolicyRule>,
+}
+
+impl PolicyEngine {
+    #[must_use]
+    pub fn new(mut rules: Vec<PolicyRule>) -> Self {
+        rules.sort_by_key(|rule| rule.priority);
+        Self { rules }
+    }
+
+    #[must_use]
+    pub fn rules(&self) -> &[PolicyRule] {
+        &self.rules
+    }
+
+    #[must_use]
+    pub fn evaluate(&self, context: &LaneContext) -> Vec<PolicyAction> {
+        evaluate(self, context)
+    }
+}
+
+#[must_use]
+pub fn evaluate(engine: &PolicyEngine, context: &LaneContext) -> Vec<PolicyAction> {
+    let mut actions = Vec::new();
+    for rule in &engine.rules {
+        if rule.matches(context) {
+            rule.action.flatten_into(&mut actions);
+        }
+    }
+    actions
+}
+
+#[cfg(test)]
+mod tests {
+    use std::time::Duration;
+
+    use super::{
+        evaluate, DiffScope, LaneBlocker, LaneContext, PolicyAction, PolicyCondition, PolicyEngine,
+        PolicyRule, ReconcileReason, ReviewStatus, STALE_BRANCH_THRESHOLD,
+    };
+
+    fn default_context() -> LaneContext {
+        LaneContext::new(
+            "lane-7",
+            0,
+            Duration::from_secs(0),
+            LaneBlocker::None,
+            ReviewStatus::Pending,
+            DiffScope::Full,
+            false,
+        )
+    }
+
+    #[test]
+    fn merge_to_dev_rule_fires_for_green_scoped_reviewed_lane() {
+        // given
+        let engine = PolicyEngine::new(vec![PolicyRule::new(
+            "merge-to-dev",
+            PolicyCondition::And(vec![
+                PolicyCondition::GreenAt { level: 2 },
+                PolicyCondition::ScopedDiff,
+                PolicyCondition::ReviewPassed,
+            ]),
+            PolicyAction::MergeToDev,
+            20,
+        )]);
+        let context = LaneContext::new(
+            "lane-7",
+            3,
+            Duration::from_secs(5),
+            LaneBlocker::None,
+            ReviewStatus::Approved,
+            DiffScope::Scoped,
+            false,
+        );
+
+        // when
+        let actions = engine.evaluate(&context);
+
+        // then
+        assert_eq!(actions, vec![PolicyAction::MergeToDev]);
+    }
+
+    #[test]
+    fn stale_branch_rule_fires_at_threshold() {
+        // given
+        let engine = PolicyEngine::new(vec![PolicyRule::new(
+            "merge-forward",
+            PolicyCondition::StaleBranch,
+            PolicyAction::MergeForward,
+            10,
+        )]);
+        let context = LaneContext::new(
+            "lane-7",
+            1,
+            STALE_BRANCH_THRESHOLD,
+            LaneBlocker::None,
+            ReviewStatus::Pending,
+            DiffScope::Full,
+            false,
+        );
+
+        // when
+        let actions = engine.evaluate(&context);
+
+        // then
+        assert_eq!(actions, vec![PolicyAction::MergeForward]);
+    }
+
+    #[test]
+    fn startup_blocked_rule_recovers_then_escalates() {
+        // given
+        let engine = PolicyEngine::new(vec![PolicyRule::new(
+            "startup-recovery",
+            PolicyCondition::StartupBlocked,
+            PolicyAction::Chain(vec![
+                PolicyAction::RecoverOnce,
+                PolicyAction::Escalate {
+                    reason: "startup remained blocked".to_string(),
+                },
+            ]),
+            15,
+        )]);
+        let context = LaneContext::new(
+            "lane-7",
+            0,
+            Duration::from_secs(0),
+            LaneBlocker::Startup,
+            ReviewStatus::Pending,
+            DiffScope::Full,
+            false,
+        );
+
+        // when
+        let actions = engine.evaluate(&context);
+
+        // then
+        assert_eq!(
+            actions,
+            vec![
+                PolicyAction::RecoverOnce,
+                PolicyAction::Escalate {
+                    reason: "startup remained blocked".to_string(),
+                },
+            ]
+        );
+    }
+
+    #[test]
+    fn completed_lane_rule_closes_out_and_cleans_up() {
+        // given
+        let engine = PolicyEngine::new(vec![PolicyRule::new(
+            "lane-closeout",
+            PolicyCondition::LaneCompleted,
+            PolicyAction::Chain(vec![
+                PolicyAction::CloseoutLane,
+                PolicyAction::CleanupSession,
+            ]),
+            30,
+        )]);
+        let context = LaneContext::new(
+            "lane-7",
+            0,
+            Duration::from_secs(0),
+            LaneBlocker::None,
+            ReviewStatus::Pending,
+            DiffScope::Full,
+            true,
+        );
+
+        // when
+        let actions = engine.evaluate(&context);
+
+        // then
+        assert_eq!(
+            actions,
+            vec![PolicyAction::CloseoutLane, PolicyAction::CleanupSession]
+        );
+    }
+
+    #[test]
+    fn matching_rules_are_returned_in_priority_order_with_stable_ties() {
+        // given
+        let engine = PolicyEngine::new(vec![
+            PolicyRule::new(
+                "late-cleanup",
+                PolicyCondition::And(vec![]),
+                PolicyAction::CleanupSession,
+                30,
+            ),
+            PolicyRule::new(
+                "first-notify",
+                PolicyCondition::And(vec![]),
+                PolicyAction::Notify {
+                    channel: "ops".to_string(),
+                },
+                10,
+            ),
+            PolicyRule::new(
+                "second-notify",
+                PolicyCondition::And(vec![]),
+                PolicyAction::Notify {
+                    channel: "review".to_string(),
+                },
+                10,
+            ),
+            PolicyRule::new(
+                "merge",
+                PolicyCondition::And(vec![]),
+                PolicyAction::MergeToDev,
+                20,
+            ),
+        ]);
+        let context = default_context();
+
+        // when
+        let actions = evaluate(&engine, &context);
+
+        // then
+        assert_eq!(
+            actions,
+            vec![
+                PolicyAction::Notify {
+                    channel: "ops".to_string(),
+                },
+                PolicyAction::Notify {
+                    channel: "review".to_string(),
+                },
+                PolicyAction::MergeToDev,
+                PolicyAction::CleanupSession,
+            ]
+        );
+    }
+
+    #[test]
+    fn combinators_handle_empty_cases_and_nested_chains() {
+        // given
+        let engine = PolicyEngine::new(vec![
+            PolicyRule::new(
+                "empty-and",
+                PolicyCondition::And(vec![]),
+                PolicyAction::Notify {
+                    channel: "orchestrator".to_string(),
+                },
+                5,
+            ),
+            PolicyRule::new(
+                "empty-or",
+                PolicyCondition::Or(vec![]),
+                PolicyAction::Block {
+                    reason: "should not fire".to_string(),
+                },
+                10,
+            ),
+            PolicyRule::new(
+                "nested",
+                PolicyCondition::Or(vec![
+                    PolicyCondition::StartupBlocked,
+                    PolicyCondition::And(vec![
+                        PolicyCondition::GreenAt { level: 2 },
+                        PolicyCondition::TimedOut {
+                            duration: Duration::from_secs(5),
+                        },
+                    ]),
+                ]),
+                PolicyAction::Chain(vec![
+                    PolicyAction::Notify {
+                        channel: "alerts".to_string(),
+                    },
+                    PolicyAction::Chain(vec![
+                        PolicyAction::MergeForward,
+                        PolicyAction::CleanupSession,
+                    ]),
+                ]),
+                15,
+            ),
+        ]);
+        let context = LaneContext::new(
+            "lane-7",
+            2,
+            Duration::from_secs(10),
+            LaneBlocker::External,
+            ReviewStatus::Pending,
+            DiffScope::Full,
+            false,
+        );
+
+        // when
+        let actions = engine.evaluate(&context);
+
+        // then
+        assert_eq!(
+            actions,
+            vec![
+                PolicyAction::Notify {
+                    channel: "orchestrator".to_string(),
+                },
+                PolicyAction::Notify {
+                    channel: "alerts".to_string(),
+                },
+                PolicyAction::MergeForward,
+                PolicyAction::CleanupSession,
+            ]
+        );
+    }
+
+    #[test]
+    fn reconciled_lane_emits_reconcile_and_cleanup() {
+        // given — a lane where branch is already merged, no PR needed, session stale
+        let engine = PolicyEngine::new(vec![
+            PolicyRule::new(
+                "reconcile-closeout",
+                PolicyCondition::LaneReconciled,
+                PolicyAction::Chain(vec![
+                    PolicyAction::Reconcile {
+                        reason: ReconcileReason::AlreadyMerged,
+                    },
+                    PolicyAction::CloseoutLane,
+                    PolicyAction::CleanupSession,
+                ]),
+                5,
+            ),
+            // This rule should NOT fire — reconciled lanes are completed but we want
+            // the more specific reconcile rule to handle them
+            PolicyRule::new(
+                "generic-closeout",
+                PolicyCondition::And(vec![
+                    PolicyCondition::LaneCompleted,
+                    // Only fire if NOT reconciled
+                    PolicyCondition::And(vec![]),
+                ]),
+                PolicyAction::CloseoutLane,
+                30,
+            ),
+        ]);
+        let context = LaneContext::reconciled("lane-9411");
+
+        // when
+        let actions = engine.evaluate(&context);
+
+        // then — reconcile rule fires first (priority 5), then generic closeout also fires
+        // because reconciled context has completed=true
+        assert_eq!(
+            actions,
+            vec![
+                PolicyAction::Reconcile {
+                    reason: ReconcileReason::AlreadyMerged,
+                },
+                PolicyAction::CloseoutLane,
+                PolicyAction::CleanupSession,
+                PolicyAction::CloseoutLane,
+            ]
+        );
+    }
+
+    #[test]
+    fn reconciled_context_has_correct_defaults() {
+        let ctx = LaneContext::reconciled("test-lane");
+        assert_eq!(ctx.lane_id, "test-lane");
+        assert!(ctx.completed);
+        assert!(ctx.reconciled);
+        assert_eq!(ctx.blocker, LaneBlocker::None);
+        assert_eq!(ctx.green_level, 0);
+    }
+
+    #[test]
+    fn non_reconciled_lane_does_not_trigger_reconcile_rule() {
+        let engine = PolicyEngine::new(vec![PolicyRule::new(
+            "reconcile-closeout",
+            PolicyCondition::LaneReconciled,
+            PolicyAction::Reconcile {
+                reason: ReconcileReason::EmptyDiff,
+            },
+            5,
+        )]);
+        // Normal completed lane — not reconciled
+        let context = LaneContext::new(
+            "lane-7",
+            0,
+            Duration::from_secs(0),
+            LaneBlocker::None,
+            ReviewStatus::Pending,
+            DiffScope::Full,
+            true,
+        );
+
+        let actions = engine.evaluate(&context);
+        assert!(actions.is_empty());
+    }
+
+    #[test]
+    fn reconcile_reason_variants_are_distinct() {
+        assert_ne!(ReconcileReason::AlreadyMerged, ReconcileReason::Superseded);
+        assert_ne!(ReconcileReason::EmptyDiff, ReconcileReason::ManualClose);
+    }
+}
@@ -4,8 +4,9 @@ use std::path::{Path, PathBuf};
 use std::process::Command;

 use crate::config::{ConfigError, ConfigLoader, RuntimeConfig};
-use lsp::LspContextEnrichment;
+use crate::git_context::GitContext;

+/// Errors raised while assembling the final system prompt.
 #[derive(Debug)]
 pub enum PromptBuildError {
    Io(std::io::Error),
@@ -35,23 +36,28 @@ impl From<ConfigError> for PromptBuildError {
    }
 }

+/// Marker separating static prompt scaffolding from dynamic runtime context.
 pub const SYSTEM_PROMPT_DYNAMIC_BOUNDARY: &str = "__SYSTEM_PROMPT_DYNAMIC_BOUNDARY__";
-pub const FRONTIER_MODEL_NAME: &str = "Opus 4.6";
+/// Human-readable default frontier model name embedded into generated prompts.
+pub const FRONTIER_MODEL_NAME: &str = "Claude Opus 4.6";
 const MAX_INSTRUCTION_FILE_CHARS: usize = 4_000;
 const MAX_TOTAL_INSTRUCTION_CHARS: usize = 12_000;

+/// Contents of an instruction file included in prompt construction.
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct ContextFile {
    pub path: PathBuf,
    pub content: String,
 }

+/// Project-local context injected into the rendered system prompt.
 #[derive(Debug, Clone, Default, PartialEq, Eq)]
 pub struct ProjectContext {
    pub cwd: PathBuf,
    pub current_date: String,
    pub git_status: Option<String>,
    pub git_diff: Option<String>,
+    pub git_context: Option<GitContext>,
    pub instruction_files: Vec<ContextFile>,
 }

@@ -67,6 +73,7 @@ impl ProjectContext {
            current_date: current_date.into(),
            git_status: None,
            git_diff: None,
+            git_context: None,
            instruction_files,
        })
    }
@@ -78,10 +85,12 @@ impl ProjectContext {
        let mut context = Self::discover(cwd, current_date)?;
        context.git_status = read_git_status(&context.cwd);
        context.git_diff = read_git_diff(&context.cwd);
+        context.git_context = GitContext::detect(&context.cwd);
        Ok(context)
    }
 }

+/// Builder for the runtime system prompt and dynamic environment sections.
 #[derive(Debug, Clone, Default, PartialEq, Eq)]
 pub struct SystemPromptBuilder {
    output_style_name: Option<String>,
@@ -131,15 +140,6 @@ impl SystemPromptBuilder {
        self
    }

-    #[must_use]
-    pub fn with_lsp_context(mut self, enrichment: &LspContextEnrichment) -> Self {
-        if !enrichment.is_empty() {
-            self.append_sections
-                .push(enrichment.render_prompt_section());
-        }
-        self
-    }
-
    #[must_use]
    pub fn build(&self) -> Vec<String> {
        let mut sections = Vec::new();
@@ -194,6 +194,7 @@ impl SystemPromptBuilder {
    }
 }

+/// Formats each item as an indented bullet for prompt sections.
 #[must_use]
 pub fn prepend_bullets(items: Vec<String>) -> Vec<String> {
    items.into_iter().map(|item| format!(" - {item}")).collect()
@@ -211,9 +212,9 @@ fn discover_instruction_files(cwd: &Path) -> std::io::Result<Vec<ContextFile>> {
    let mut files = Vec::new();
    for dir in directories {
        for candidate in [
-            dir.join("CLAW.md"),
-            dir.join("CLAW.local.md"),
-            dir.join(".claw").join("CLAW.md"),
+            dir.join("CLAUDE.md"),
+            dir.join("CLAUDE.local.md"),
+            dir.join(".claw").join("CLAUDE.md"),
            dir.join(".claw").join("instructions.md"),
        ] {
            push_context_file(&mut files, candidate)?;
@@ -292,7 +293,7 @@ fn render_project_context(project_context: &ProjectContext) -> String {
    ];
    if !project_context.instruction_files.is_empty() {
        bullets.push(format!(
-            "Claw instruction files discovered: {}.",
+            "Claude instruction files discovered: {}.",
            project_context.instruction_files.len()
        ));
    }
@@ -302,16 +303,32 @@ fn render_project_context(project_context: &ProjectContext) -> String {
        lines.push("Git status snapshot:".to_string());
        lines.push(status.clone());
    }
+    if let Some(ref gc) = project_context.git_context {
+        if !gc.recent_commits.is_empty() {
+            lines.push(String::new());
+            lines.push("Recent commits (last 5):".to_string());
+            for c in &gc.recent_commits {
+                lines.push(format!("  {} {}", c.hash, c.subject));
+            }
+        }
+    }
    if let Some(diff) = &project_context.git_diff {
        lines.push(String::new());
        lines.push("Git diff snapshot:".to_string());
        lines.push(diff.clone());
    }
+    if let Some(git_context) = &project_context.git_context {
+        let rendered = git_context.render();
+        if !rendered.is_empty() {
+            lines.push(String::new());
+            lines.push(rendered);
+        }
+    }
    lines.join("\n")
 }

 fn render_instruction_files(files: &[ContextFile]) -> String {
-    let mut sections = vec!["# Claw instructions".to_string()];
+    let mut sections = vec!["# Claude instructions".to_string()];
    let mut remaining_chars = MAX_TOTAL_INSTRUCTION_CHARS;
    for file in files {
        if remaining_chars == 0 {
@@ -411,6 +428,7 @@ fn collapse_blank_lines(content: &str) -> String {
    result
 }

+/// Loads config and project context, then renders the system prompt text.
 pub fn load_system_prompt(
    cwd: impl Into<PathBuf>,
    current_date: impl Into<String>,
@@ -523,24 +541,31 @@ mod tests {
        crate::test_env_lock()
    }

+    fn ensure_valid_cwd() {
+        if std::env::current_dir().is_err() {
+            std::env::set_current_dir(env!("CARGO_MANIFEST_DIR"))
+                .expect("test cwd should be recoverable");
+        }
+    }
+
    #[test]
    fn discovers_instruction_files_from_ancestor_chain() {
        let root = temp_dir();
        let nested = root.join("apps").join("api");
        fs::create_dir_all(nested.join(".claw")).expect("nested claw dir");
-        fs::write(root.join("CLAW.md"), "root instructions").expect("write root instructions");
-        fs::write(root.join("CLAW.local.md"), "local instructions")
+        fs::write(root.join("CLAUDE.md"), "root instructions").expect("write root instructions");
+        fs::write(root.join("CLAUDE.local.md"), "local instructions")
            .expect("write local instructions");
        fs::create_dir_all(root.join("apps")).expect("apps dir");
        fs::create_dir_all(root.join("apps").join(".claw")).expect("apps claw dir");
-        fs::write(root.join("apps").join("CLAW.md"), "apps instructions")
+        fs::write(root.join("apps").join("CLAUDE.md"), "apps instructions")
            .expect("write apps instructions");
        fs::write(
            root.join("apps").join(".claw").join("instructions.md"),
-            "apps dot claw instructions",
+            "apps dot claude instructions",
        )
-        .expect("write apps dot claw instructions");
-        fs::write(nested.join(".claw").join("CLAW.md"), "nested rules")
+        .expect("write apps dot claude instructions");
+        fs::write(nested.join(".claw").join("CLAUDE.md"), "nested rules")
            .expect("write nested rules");
        fs::write(
            nested.join(".claw").join("instructions.md"),
@@ -561,7 +586,7 @@ mod tests {
                "root instructions",
                "local instructions",
                "apps instructions",
-                "apps dot claw instructions",
+                "apps dot claude instructions",
                "nested rules",
                "nested instructions"
            ]
@@ -574,8 +599,8 @@ mod tests {
        let root = temp_dir();
        let nested = root.join("apps").join("api");
        fs::create_dir_all(&nested).expect("nested dir");
-        fs::write(root.join("CLAW.md"), "same rules\n\n").expect("write root");
-        fs::write(nested.join("CLAW.md"), "same rules\n").expect("write nested");
+        fs::write(root.join("CLAUDE.md"), "same rules\n\n").expect("write root");
+        fs::write(nested.join("CLAUDE.md"), "same rules\n").expect("write nested");

        let context = ProjectContext::discover(&nested, "2026-03-31").expect("context should load");
        assert_eq!(context.instruction_files.len(), 1);
@@ -603,14 +628,15 @@ mod tests {
    #[test]
    fn displays_context_paths_compactly() {
        assert_eq!(
-            display_context_path(Path::new("/tmp/project/.claw/CLAW.md")),
-            "CLAW.md"
+            display_context_path(Path::new("/tmp/project/.claw/CLAUDE.md")),
+            "CLAUDE.md"
        );
    }

    #[test]
    fn discover_with_git_includes_status_snapshot() {
        let _guard = env_lock();
+        ensure_valid_cwd();
        let root = temp_dir();
        fs::create_dir_all(&root).expect("root dir");
        std::process::Command::new("git")
@@ -618,7 +644,7 @@ mod tests {
            .current_dir(&root)
            .status()
            .expect("git init should run");
-        fs::write(root.join("CLAW.md"), "rules").expect("write instructions");
+        fs::write(root.join("CLAUDE.md"), "rules").expect("write instructions");
        fs::write(root.join("tracked.txt"), "hello").expect("write tracked file");

        let context =
@@ -626,16 +652,99 @@ mod tests {

        let status = context.git_status.expect("git status should be present");
        assert!(status.contains("## No commits yet on") || status.contains("## "));
-        assert!(status.contains("?? CLAW.md"));
+        assert!(status.contains("?? CLAUDE.md"));
        assert!(status.contains("?? tracked.txt"));
        assert!(context.git_diff.is_none());

        fs::remove_dir_all(root).expect("cleanup temp dir");
    }

+    #[test]
+    fn discover_with_git_includes_recent_commits_and_renders_them() {
+        // given: a git repo with three commits and a current branch
+        let _guard = env_lock();
+        ensure_valid_cwd();
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("root dir");
+        std::process::Command::new("git")
+            .args(["init", "--quiet", "-b", "main"])
+            .current_dir(&root)
+            .status()
+            .expect("git init should run");
+        std::process::Command::new("git")
+            .args(["config", "user.email", "tests@example.com"])
+            .current_dir(&root)
+            .status()
+            .expect("git config email should run");
+        std::process::Command::new("git")
+            .args(["config", "user.name", "Runtime Prompt Tests"])
+            .current_dir(&root)
+            .status()
+            .expect("git config name should run");
+        for (file, message) in [
+            ("a.txt", "first commit"),
+            ("b.txt", "second commit"),
+            ("c.txt", "third commit"),
+        ] {
+            fs::write(root.join(file), "x\n").expect("write commit file");
+            std::process::Command::new("git")
+                .args(["add", file])
+                .current_dir(&root)
+                .status()
+                .expect("git add should run");
+            std::process::Command::new("git")
+                .args(["commit", "-m", message, "--quiet"])
+                .current_dir(&root)
+                .status()
+                .expect("git commit should run");
+        }
+        fs::write(root.join("d.txt"), "staged\n").expect("write staged file");
+        std::process::Command::new("git")
+            .args(["add", "d.txt"])
+            .current_dir(&root)
+            .status()
+            .expect("git add staged should run");
+
+        // when: discovering project context with git auto-include
+        let context =
+            ProjectContext::discover_with_git(&root, "2026-03-31").expect("context should load");
+        let rendered = SystemPromptBuilder::new()
+            .with_os("linux", "6.8")
+            .with_project_context(context.clone())
+            .render();
+
+        // then: branch, recent commits and staged files are present in context
+        let gc = context
+            .git_context
+            .as_ref()
+            .expect("git context should be present");
+        let commits: String = gc
+            .recent_commits
+            .iter()
+            .map(|c| c.subject.clone())
+            .collect::<Vec<_>>()
+            .join("\n");
+        assert!(commits.contains("first commit"));
+        assert!(commits.contains("second commit"));
+        assert!(commits.contains("third commit"));
+        assert_eq!(gc.recent_commits.len(), 3);
+
+        let status = context.git_status.as_deref().expect("status snapshot");
+        assert!(status.contains("## main"));
+        assert!(status.contains("A  d.txt"));
+
+        assert!(rendered.contains("Recent commits (last 5):"));
+        assert!(rendered.contains("first commit"));
+        assert!(rendered.contains("Git status snapshot:"));
+        assert!(rendered.contains("## main"));
+
+        fs::remove_dir_all(root).expect("cleanup temp dir");
+    }
+
    #[test]
    fn discover_with_git_includes_diff_snapshot_for_tracked_changes() {
        let _guard = env_lock();
+        ensure_valid_cwd();
        let root = temp_dir();
        fs::create_dir_all(&root).expect("root dir");
        std::process::Command::new("git")
@@ -677,10 +786,10 @@ mod tests {
    }

    #[test]
-    fn load_system_prompt_reads_claw_files_and_config() {
+    fn load_system_prompt_reads_claude_files_and_config() {
        let root = temp_dir();
        fs::create_dir_all(root.join(".claw")).expect("claw dir");
-        fs::write(root.join("CLAW.md"), "Project rules").expect("write instructions");
+        fs::write(root.join("CLAUDE.md"), "Project rules").expect("write instructions");
        fs::write(
            root.join(".claw").join("settings.json"),
            r#"{"permissionMode":"acceptEdits"}"#,
@@ -688,6 +797,7 @@ mod tests {
        .expect("write settings");

        let _guard = env_lock();
+        ensure_valid_cwd();
        let previous = std::env::current_dir().expect("cwd");
        let original_home = std::env::var("HOME").ok();
        let original_claw_home = std::env::var("CLAW_CONFIG_HOME").ok();
@@ -719,10 +829,10 @@ mod tests {
    }

    #[test]
-    fn renders_claw_code_style_sections_with_project_context() {
+    fn renders_claude_code_style_sections_with_project_context() {
        let root = temp_dir();
        fs::create_dir_all(root.join(".claw")).expect("claw dir");
-        fs::write(root.join("CLAW.md"), "Project rules").expect("write CLAW.md");
+        fs::write(root.join("CLAUDE.md"), "Project rules").expect("write CLAUDE.md");
        fs::write(
            root.join(".claw").join("settings.json"),
            r#"{"permissionMode":"acceptEdits"}"#,
@@ -743,7 +853,7 @@ mod tests {

        assert!(prompt.contains("# System"));
        assert!(prompt.contains("# Project context"));
-        assert!(prompt.contains("# Claw instructions"));
+        assert!(prompt.contains("# Claude instructions"));
        assert!(prompt.contains("Project rules"));
        assert!(prompt.contains("permissionMode"));
        assert!(prompt.contains(SYSTEM_PROMPT_DYNAMIC_BOUNDARY));
@@ -760,7 +870,7 @@ mod tests {
    }

    #[test]
-    fn discovers_dot_claw_instructions_markdown() {
+    fn discovers_dot_claude_instructions_markdown() {
        let root = temp_dir();
        let nested = root.join("apps").join("api");
        fs::create_dir_all(nested.join(".claw")).expect("nested claw dir");
@@ -785,10 +895,10 @@ mod tests {
    #[test]
    fn renders_instruction_file_metadata() {
        let rendered = render_instruction_files(&[ContextFile {
-            path: PathBuf::from("/tmp/project/CLAW.md"),
+            path: PathBuf::from("/tmp/project/CLAUDE.md"),
            content: "Project rules".to_string(),
        }]);
-        assert!(rendered.contains("# Claw instructions"));
+        assert!(rendered.contains("# Claude instructions"));
        assert!(rendered.contains("scope: /tmp/project"));
        assert!(rendered.contains("Project rules"));
    }
@@ -0,0 +1,631 @@
+#![allow(clippy::cast_possible_truncation, clippy::uninlined_format_args)]
+//! Recovery recipes for common failure scenarios.
+//!
+//! Encodes known automatic recoveries for the six failure scenarios
+//! listed in ROADMAP item 8, and enforces one automatic recovery
+//! attempt before escalation. Each attempt is emitted as a structured
+//! recovery event.
+
+use std::collections::HashMap;
+
+use serde::{Deserialize, Serialize};
+
+use crate::worker_boot::WorkerFailureKind;
+
+/// The six failure scenarios that have known recovery recipes.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum FailureScenario {
+    TrustPromptUnresolved,
+    PromptMisdelivery,
+    StaleBranch,
+    CompileRedCrossCrate,
+    McpHandshakeFailure,
+    PartialPluginStartup,
+    ProviderFailure,
+}
+
+impl FailureScenario {
+    /// Returns all known failure scenarios.
+    #[must_use]
+    pub fn all() -> &'static [FailureScenario] {
+        &[
+            Self::TrustPromptUnresolved,
+            Self::PromptMisdelivery,
+            Self::StaleBranch,
+            Self::CompileRedCrossCrate,
+            Self::McpHandshakeFailure,
+            Self::PartialPluginStartup,
+            Self::ProviderFailure,
+        ]
+    }
+
+    /// Map a `WorkerFailureKind` to the corresponding `FailureScenario`.
+    /// This is the bridge that lets recovery policy consume worker boot events.
+    #[must_use]
+    pub fn from_worker_failure_kind(kind: WorkerFailureKind) -> Self {
+        match kind {
+            WorkerFailureKind::TrustGate => Self::TrustPromptUnresolved,
+            WorkerFailureKind::PromptDelivery => Self::PromptMisdelivery,
+            WorkerFailureKind::Protocol => Self::McpHandshakeFailure,
+            WorkerFailureKind::Provider => Self::ProviderFailure,
+        }
+    }
+}
+
+impl std::fmt::Display for FailureScenario {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::TrustPromptUnresolved => write!(f, "trust_prompt_unresolved"),
+            Self::PromptMisdelivery => write!(f, "prompt_misdelivery"),
+            Self::StaleBranch => write!(f, "stale_branch"),
+            Self::CompileRedCrossCrate => write!(f, "compile_red_cross_crate"),
+            Self::McpHandshakeFailure => write!(f, "mcp_handshake_failure"),
+            Self::PartialPluginStartup => write!(f, "partial_plugin_startup"),
+            Self::ProviderFailure => write!(f, "provider_failure"),
+        }
+    }
+}
+
+/// Individual step that can be executed as part of a recovery recipe.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum RecoveryStep {
+    AcceptTrustPrompt,
+    RedirectPromptToAgent,
+    RebaseBranch,
+    CleanBuild,
+    RetryMcpHandshake { timeout: u64 },
+    RestartPlugin { name: String },
+    RestartWorker,
+    EscalateToHuman { reason: String },
+}
+
+/// Policy governing what happens when automatic recovery is exhausted.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum EscalationPolicy {
+    AlertHuman,
+    LogAndContinue,
+    Abort,
+}
+
+/// A recovery recipe encodes the sequence of steps to attempt for a
+/// given failure scenario, along with the maximum number of automatic
+/// attempts and the escalation policy.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct RecoveryRecipe {
+    pub scenario: FailureScenario,
+    pub steps: Vec<RecoveryStep>,
+    pub max_attempts: u32,
+    pub escalation_policy: EscalationPolicy,
+}
+
+/// Outcome of a recovery attempt.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum RecoveryResult {
+    Recovered {
+        steps_taken: u32,
+    },
+    PartialRecovery {
+        recovered: Vec<RecoveryStep>,
+        remaining: Vec<RecoveryStep>,
+    },
+    EscalationRequired {
+        reason: String,
+    },
+}
+
+/// Structured event emitted during recovery.
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum RecoveryEvent {
+    RecoveryAttempted {
+        scenario: FailureScenario,
+        recipe: RecoveryRecipe,
+        result: RecoveryResult,
+    },
+    RecoverySucceeded,
+    RecoveryFailed,
+    Escalated,
+}
+
+/// Minimal context for tracking recovery state and emitting events.
+///
+/// Holds per-scenario attempt counts, a structured event log, and an
+/// optional simulation knob for controlling step outcomes during tests.
+#[derive(Debug, Clone, Default)]
+pub struct RecoveryContext {
+    attempts: HashMap<FailureScenario, u32>,
+    events: Vec<RecoveryEvent>,
+    /// Optional step index at which simulated execution fails.
+    /// `None` means all steps succeed.
+    fail_at_step: Option<usize>,
+}
+
+impl RecoveryContext {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    /// Configure a step index at which simulated execution will fail.
+    #[must_use]
+    pub fn with_fail_at_step(mut self, index: usize) -> Self {
+        self.fail_at_step = Some(index);
+        self
+    }
+
+    /// Returns the structured event log populated during recovery.
+    #[must_use]
+    pub fn events(&self) -> &[RecoveryEvent] {
+        &self.events
+    }
+
+    /// Returns the number of recovery attempts made for a scenario.
+    #[must_use]
+    pub fn attempt_count(&self, scenario: &FailureScenario) -> u32 {
+        self.attempts.get(scenario).copied().unwrap_or(0)
+    }
+}
+
+/// Returns the known recovery recipe for the given failure scenario.
+#[must_use]
+pub fn recipe_for(scenario: &FailureScenario) -> RecoveryRecipe {
+    match scenario {
+        FailureScenario::TrustPromptUnresolved => RecoveryRecipe {
+            scenario: *scenario,
+            steps: vec![RecoveryStep::AcceptTrustPrompt],
+            max_attempts: 1,
+            escalation_policy: EscalationPolicy::AlertHuman,
+        },
+        FailureScenario::PromptMisdelivery => RecoveryRecipe {
+            scenario: *scenario,
+            steps: vec![RecoveryStep::RedirectPromptToAgent],
+            max_attempts: 1,
+            escalation_policy: EscalationPolicy::AlertHuman,
+        },
+        FailureScenario::StaleBranch => RecoveryRecipe {
+            scenario: *scenario,
+            steps: vec![RecoveryStep::RebaseBranch, RecoveryStep::CleanBuild],
+            max_attempts: 1,
+            escalation_policy: EscalationPolicy::AlertHuman,
+        },
+        FailureScenario::CompileRedCrossCrate => RecoveryRecipe {
+            scenario: *scenario,
+            steps: vec![RecoveryStep::CleanBuild],
+            max_attempts: 1,
+            escalation_policy: EscalationPolicy::AlertHuman,
+        },
+        FailureScenario::McpHandshakeFailure => RecoveryRecipe {
+            scenario: *scenario,
+            steps: vec![RecoveryStep::RetryMcpHandshake { timeout: 5000 }],
+            max_attempts: 1,
+            escalation_policy: EscalationPolicy::Abort,
+        },
+        FailureScenario::PartialPluginStartup => RecoveryRecipe {
+            scenario: *scenario,
+            steps: vec![
+                RecoveryStep::RestartPlugin {
+                    name: "stalled".to_string(),
+                },
+                RecoveryStep::RetryMcpHandshake { timeout: 3000 },
+            ],
+            max_attempts: 1,
+            escalation_policy: EscalationPolicy::LogAndContinue,
+        },
+        FailureScenario::ProviderFailure => RecoveryRecipe {
+            scenario: *scenario,
+            steps: vec![RecoveryStep::RestartWorker],
+            max_attempts: 1,
+            escalation_policy: EscalationPolicy::AlertHuman,
+        },
+    }
+}
+
+/// Attempts automatic recovery for the given failure scenario.
+///
+/// Looks up the recipe, enforces the one-attempt-before-escalation
+/// policy, simulates step execution (controlled by the context), and
+/// emits structured [`RecoveryEvent`]s for every attempt.
+pub fn attempt_recovery(scenario: &FailureScenario, ctx: &mut RecoveryContext) -> RecoveryResult {
+    let recipe = recipe_for(scenario);
+    let attempt_count = ctx.attempts.entry(*scenario).or_insert(0);
+
+    // Enforce one automatic recovery attempt before escalation.
+    if *attempt_count >= recipe.max_attempts {
+        let result = RecoveryResult::EscalationRequired {
+            reason: format!(
+                "max recovery attempts ({}) exceeded for {}",
+                recipe.max_attempts, scenario
+            ),
+        };
+        ctx.events.push(RecoveryEvent::RecoveryAttempted {
+            scenario: *scenario,
+            recipe,
+            result: result.clone(),
+        });
+        ctx.events.push(RecoveryEvent::Escalated);
+        return result;
+    }
+
+    *attempt_count += 1;
+
+    // Execute steps, honoring the optional fail_at_step simulation.
+    let fail_index = ctx.fail_at_step;
+    let mut executed = Vec::new();
+    let mut failed = false;
+
+    for (i, step) in recipe.steps.iter().enumerate() {
+        if fail_index == Some(i) {
+            failed = true;
+            break;
+        }
+        executed.push(step.clone());
+    }
+
+    let result = if failed {
+        let remaining: Vec<RecoveryStep> = recipe.steps[executed.len()..].to_vec();
+        if executed.is_empty() {
+            RecoveryResult::EscalationRequired {
+                reason: format!("recovery failed at first step for {}", scenario),
+            }
+        } else {
+            RecoveryResult::PartialRecovery {
+                recovered: executed,
+                remaining,
+            }
+        }
+    } else {
+        RecoveryResult::Recovered {
+            steps_taken: recipe.steps.len() as u32,
+        }
+    };
+
+    // Emit the attempt as structured event data.
+    ctx.events.push(RecoveryEvent::RecoveryAttempted {
+        scenario: *scenario,
+        recipe,
+        result: result.clone(),
+    });
+
+    match &result {
+        RecoveryResult::Recovered { .. } => {
+            ctx.events.push(RecoveryEvent::RecoverySucceeded);
+        }
+        RecoveryResult::PartialRecovery { .. } => {
+            ctx.events.push(RecoveryEvent::RecoveryFailed);
+        }
+        RecoveryResult::EscalationRequired { .. } => {
+            ctx.events.push(RecoveryEvent::Escalated);
+        }
+    }
+
+    result
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn each_scenario_has_a_matching_recipe() {
+        // given
+        let scenarios = FailureScenario::all();
+
+        // when / then
+        for scenario in scenarios {
+            let recipe = recipe_for(scenario);
+            assert_eq!(
+                recipe.scenario, *scenario,
+                "recipe scenario should match requested scenario"
+            );
+            assert!(
+                !recipe.steps.is_empty(),
+                "recipe for {} should have at least one step",
+                scenario
+            );
+            assert!(
+                recipe.max_attempts >= 1,
+                "recipe for {} should allow at least one attempt",
+                scenario
+            );
+        }
+    }
+
+    #[test]
+    fn successful_recovery_returns_recovered_and_emits_events() {
+        // given
+        let mut ctx = RecoveryContext::new();
+        let scenario = FailureScenario::TrustPromptUnresolved;
+
+        // when
+        let result = attempt_recovery(&scenario, &mut ctx);
+
+        // then
+        assert_eq!(result, RecoveryResult::Recovered { steps_taken: 1 });
+        assert_eq!(ctx.events().len(), 2);
+        assert!(matches!(
+            &ctx.events()[0],
+            RecoveryEvent::RecoveryAttempted {
+                scenario: s,
+                result: r,
+                ..
+            } if *s == FailureScenario::TrustPromptUnresolved
+              && matches!(r, RecoveryResult::Recovered { steps_taken: 1 })
+        ));
+        assert_eq!(ctx.events()[1], RecoveryEvent::RecoverySucceeded);
+    }
+
+    #[test]
+    fn escalation_after_max_attempts_exceeded() {
+        // given
+        let mut ctx = RecoveryContext::new();
+        let scenario = FailureScenario::PromptMisdelivery;
+
+        // when — first attempt succeeds
+        let first = attempt_recovery(&scenario, &mut ctx);
+        assert!(matches!(first, RecoveryResult::Recovered { .. }));
+
+        // when — second attempt should escalate
+        let second = attempt_recovery(&scenario, &mut ctx);
+
+        // then
+        assert!(
+            matches!(
+                &second,
+                RecoveryResult::EscalationRequired { reason }
+                    if reason.contains("max recovery attempts")
+            ),
+            "second attempt should require escalation, got: {second:?}"
+        );
+        assert_eq!(ctx.attempt_count(&scenario), 1);
+        assert!(ctx
+            .events()
+            .iter()
+            .any(|e| matches!(e, RecoveryEvent::Escalated)));
+    }
+
+    #[test]
+    fn partial_recovery_when_step_fails_midway() {
+        // given — PartialPluginStartup has two steps; fail at step index 1
+        let mut ctx = RecoveryContext::new().with_fail_at_step(1);
+        let scenario = FailureScenario::PartialPluginStartup;
+
+        // when
+        let result = attempt_recovery(&scenario, &mut ctx);
+
+        // then
+        match &result {
+            RecoveryResult::PartialRecovery {
+                recovered,
+                remaining,
+            } => {
+                assert_eq!(recovered.len(), 1, "one step should have succeeded");
+                assert_eq!(remaining.len(), 1, "one step should remain");
+                assert!(matches!(recovered[0], RecoveryStep::RestartPlugin { .. }));
+                assert!(matches!(
+                    remaining[0],
+                    RecoveryStep::RetryMcpHandshake { .. }
+                ));
+            }
+            other => panic!("expected PartialRecovery, got {other:?}"),
+        }
+        assert!(ctx
+            .events()
+            .iter()
+            .any(|e| matches!(e, RecoveryEvent::RecoveryFailed)));
+    }
+
+    #[test]
+    fn first_step_failure_escalates_immediately() {
+        // given — fail at step index 0
+        let mut ctx = RecoveryContext::new().with_fail_at_step(0);
+        let scenario = FailureScenario::CompileRedCrossCrate;
+
+        // when
+        let result = attempt_recovery(&scenario, &mut ctx);
+
+        // then
+        assert!(
+            matches!(
+                &result,
+                RecoveryResult::EscalationRequired { reason }
+                    if reason.contains("failed at first step")
+            ),
+            "zero-step failure should escalate, got: {result:?}"
+        );
+        assert!(ctx
+            .events()
+            .iter()
+            .any(|e| matches!(e, RecoveryEvent::Escalated)));
+    }
+
+    #[test]
+    fn emitted_events_include_structured_attempt_data() {
+        // given
+        let mut ctx = RecoveryContext::new();
+        let scenario = FailureScenario::McpHandshakeFailure;
+
+        // when
+        let _ = attempt_recovery(&scenario, &mut ctx);
+
+        // then — verify the RecoveryAttempted event carries full context
+        let attempted = ctx
+            .events()
+            .iter()
+            .find(|e| matches!(e, RecoveryEvent::RecoveryAttempted { .. }))
+            .expect("should have emitted RecoveryAttempted event");
+
+        match attempted {
+            RecoveryEvent::RecoveryAttempted {
+                scenario: s,
+                recipe,
+                result,
+            } => {
+                assert_eq!(*s, scenario);
+                assert_eq!(recipe.scenario, scenario);
+                assert!(!recipe.steps.is_empty());
+                assert!(matches!(result, RecoveryResult::Recovered { .. }));
+            }
+            _ => unreachable!(),
+        }
+
+        // Verify the event is serializable as structured JSON
+        let json = serde_json::to_string(&ctx.events()[0])
+            .expect("recovery event should be serializable to JSON");
+        assert!(
+            json.contains("mcp_handshake_failure"),
+            "serialized event should contain scenario name"
+        );
+    }
+
+    #[test]
+    fn recovery_context_tracks_attempts_per_scenario() {
+        // given
+        let mut ctx = RecoveryContext::new();
+
+        // when
+        assert_eq!(ctx.attempt_count(&FailureScenario::StaleBranch), 0);
+        attempt_recovery(&FailureScenario::StaleBranch, &mut ctx);
+
+        // then
+        assert_eq!(ctx.attempt_count(&FailureScenario::StaleBranch), 1);
+        assert_eq!(ctx.attempt_count(&FailureScenario::PromptMisdelivery), 0);
+    }
+
+    #[test]
+    fn stale_branch_recipe_has_rebase_then_clean_build() {
+        // given
+        let recipe = recipe_for(&FailureScenario::StaleBranch);
+
+        // then
+        assert_eq!(recipe.steps.len(), 2);
+        assert_eq!(recipe.steps[0], RecoveryStep::RebaseBranch);
+        assert_eq!(recipe.steps[1], RecoveryStep::CleanBuild);
+    }
+
+    #[test]
+    fn partial_plugin_startup_recipe_has_restart_then_handshake() {
+        // given
+        let recipe = recipe_for(&FailureScenario::PartialPluginStartup);
+
+        // then
+        assert_eq!(recipe.steps.len(), 2);
+        assert!(matches!(
+            recipe.steps[0],
+            RecoveryStep::RestartPlugin { .. }
+        ));
+        assert!(matches!(
+            recipe.steps[1],
+            RecoveryStep::RetryMcpHandshake { timeout: 3000 }
+        ));
+        assert_eq!(recipe.escalation_policy, EscalationPolicy::LogAndContinue);
+    }
+
+    #[test]
+    fn failure_scenario_display_all_variants() {
+        // given
+        let cases = [
+            (
+                FailureScenario::TrustPromptUnresolved,
+                "trust_prompt_unresolved",
+            ),
+            (FailureScenario::PromptMisdelivery, "prompt_misdelivery"),
+            (FailureScenario::StaleBranch, "stale_branch"),
+            (
+                FailureScenario::CompileRedCrossCrate,
+                "compile_red_cross_crate",
+            ),
+            (
+                FailureScenario::McpHandshakeFailure,
+                "mcp_handshake_failure",
+            ),
+            (
+                FailureScenario::PartialPluginStartup,
+                "partial_plugin_startup",
+            ),
+        ];
+
+        // when / then
+        for (scenario, expected) in &cases {
+            assert_eq!(scenario.to_string(), *expected);
+        }
+    }
+
+    #[test]
+    fn multi_step_success_reports_correct_steps_taken() {
+        // given — StaleBranch has 2 steps, no simulated failure
+        let mut ctx = RecoveryContext::new();
+        let scenario = FailureScenario::StaleBranch;
+
+        // when
+        let result = attempt_recovery(&scenario, &mut ctx);
+
+        // then
+        assert_eq!(result, RecoveryResult::Recovered { steps_taken: 2 });
+    }
+
+    #[test]
+    fn mcp_handshake_recipe_uses_abort_escalation_policy() {
+        // given
+        let recipe = recipe_for(&FailureScenario::McpHandshakeFailure);
+
+        // then
+        assert_eq!(recipe.escalation_policy, EscalationPolicy::Abort);
+        assert_eq!(recipe.max_attempts, 1);
+    }
+
+    #[test]
+    fn worker_failure_kind_maps_to_failure_scenario() {
+        // given / when / then — verify the bridge is correct
+        assert_eq!(
+            FailureScenario::from_worker_failure_kind(WorkerFailureKind::TrustGate),
+            FailureScenario::TrustPromptUnresolved,
+        );
+        assert_eq!(
+            FailureScenario::from_worker_failure_kind(WorkerFailureKind::PromptDelivery),
+            FailureScenario::PromptMisdelivery,
+        );
+        assert_eq!(
+            FailureScenario::from_worker_failure_kind(WorkerFailureKind::Protocol),
+            FailureScenario::McpHandshakeFailure,
+        );
+        assert_eq!(
+            FailureScenario::from_worker_failure_kind(WorkerFailureKind::Provider),
+            FailureScenario::ProviderFailure,
+        );
+    }
+
+    #[test]
+    fn provider_failure_recipe_uses_restart_worker_step() {
+        // given
+        let recipe = recipe_for(&FailureScenario::ProviderFailure);
+
+        // then
+        assert_eq!(recipe.scenario, FailureScenario::ProviderFailure);
+        assert!(recipe.steps.contains(&RecoveryStep::RestartWorker));
+        assert_eq!(recipe.escalation_policy, EscalationPolicy::AlertHuman);
+        assert_eq!(recipe.max_attempts, 1);
+    }
+
+    #[test]
+    fn provider_failure_recovery_attempt_succeeds_then_escalates() {
+        // given
+        let mut ctx = RecoveryContext::new();
+        let scenario = FailureScenario::ProviderFailure;
+
+        // when — first attempt
+        let first = attempt_recovery(&scenario, &mut ctx);
+        assert!(matches!(first, RecoveryResult::Recovered { .. }));
+
+        // when — second attempt should escalate (max_attempts=1)
+        let second = attempt_recovery(&scenario, &mut ctx);
+        assert!(matches!(second, RecoveryResult::EscalationRequired { .. }));
+        assert!(ctx
+            .events()
+            .iter()
+            .any(|e| matches!(e, RecoveryEvent::Escalated)));
+    }
+}
@@ -72,9 +72,9 @@ impl RemoteSessionContext {
    #[must_use]
    pub fn from_env_map(env_map: &BTreeMap<String, String>) -> Self {
        Self {
-            enabled: env_truthy(env_map.get("CLAW_CODE_REMOTE")),
+            enabled: env_truthy(env_map.get("CLAUDE_CODE_REMOTE")),
            session_id: env_map
-                .get("CLAW_CODE_REMOTE_SESSION_ID")
+                .get("CLAUDE_CODE_REMOTE_SESSION_ID")
                .filter(|value| !value.is_empty())
                .cloned(),
            base_url: env_map
@@ -272,9 +272,9 @@ mod tests {
    #[test]
    fn remote_context_reads_env_state() {
        let env = BTreeMap::from([
-            ("CLAW_CODE_REMOTE".to_string(), "true".to_string()),
+            ("CLAUDE_CODE_REMOTE".to_string(), "true".to_string()),
            (
-                "CLAW_CODE_REMOTE_SESSION_ID".to_string(),
+                "CLAUDE_CODE_REMOTE_SESSION_ID".to_string(),
                "session-123".to_string(),
            ),
            (
@@ -291,7 +291,7 @@ mod tests {
    #[test]
    fn bootstrap_fails_open_when_token_or_session_is_missing() {
        let env = BTreeMap::from([
-            ("CLAW_CODE_REMOTE".to_string(), "1".to_string()),
+            ("CLAUDE_CODE_REMOTE".to_string(), "1".to_string()),
            ("CCR_UPSTREAM_PROXY_ENABLED".to_string(), "true".to_string()),
        ]);
        let bootstrap = UpstreamProxyBootstrap::from_env_map(&env);
@@ -307,10 +307,10 @@ mod tests {
        fs::write(&token_path, "secret-token\n").expect("write token");

        let env = BTreeMap::from([
-            ("CLAW_CODE_REMOTE".to_string(), "1".to_string()),
+            ("CLAUDE_CODE_REMOTE".to_string(), "1".to_string()),
            ("CCR_UPSTREAM_PROXY_ENABLED".to_string(), "true".to_string()),
            (
-                "CLAW_CODE_REMOTE_SESSION_ID".to_string(),
+                "CLAUDE_CODE_REMOTE_SESSION_ID".to_string(),
                "session-123".to_string(),
            ),
            (
@@ -107,23 +107,11 @@ impl SandboxConfig {

 #[must_use]
 pub fn detect_container_environment() -> ContainerEnvironment {
-    let proc_1_cgroup = if cfg!(target_os = "linux") {
-        fs::read_to_string("/proc/1/cgroup").ok()
-    } else {
-        None
-    };
+    let proc_1_cgroup = fs::read_to_string("/proc/1/cgroup").ok();
    detect_container_environment_from(SandboxDetectionInputs {
        env_pairs: env::vars().collect(),
-        dockerenv_exists: if cfg!(target_os = "linux") {
-            Path::new("/.dockerenv").exists()
-        } else {
-            false
-        },
-        containerenv_exists: if cfg!(target_os = "linux") {
-            Path::new("/run/.containerenv").exists()
-        } else {
-            false
-        },
+        dockerenv_exists: Path::new("/.dockerenv").exists(),
+        containerenv_exists: Path::new("/run/.containerenv").exists(),
        proc_1_cgroup: proc_1_cgroup.as_deref(),
    })
 }
@@ -173,7 +161,7 @@ pub fn resolve_sandbox_status(config: &SandboxConfig, cwd: &Path) -> SandboxStat
 #[must_use]
 pub fn resolve_sandbox_status_for_request(request: &SandboxRequest, cwd: &Path) -> SandboxStatus {
    let container = detect_container_environment();
-    let namespace_supported = cfg!(target_os = "linux") && command_exists("unshare");
+    let namespace_supported = cfg!(target_os = "linux") && unshare_user_namespace_works();
    let network_supported = namespace_supported;
    let filesystem_active =
        request.enabled && request.filesystem_mode != FilesystemIsolationMode::Off;
@@ -294,6 +282,27 @@ fn command_exists(command: &str) -> bool {
        .is_some_and(|paths| env::split_paths(&paths).any(|path| path.join(command).exists()))
 }

+/// Check whether `unshare --user` actually works on this system.
+/// On some CI environments (e.g. GitHub Actions), the binary exists but
+/// user namespaces are restricted, causing silent failures.
+fn unshare_user_namespace_works() -> bool {
+    use std::sync::OnceLock;
+    static RESULT: OnceLock<bool> = OnceLock::new();
+    *RESULT.get_or_init(|| {
+        if !command_exists("unshare") {
+            return false;
+        }
+        std::process::Command::new("unshare")
+            .args(["--user", "--map-root-user", "true"])
+            .stdin(std::process::Stdio::null())
+            .stdout(std::process::Stdio::null())
+            .stderr(std::process::Stdio::null())
+            .status()
+            .map(|s| s.success())
+            .unwrap_or(false)
+    })
+}
+
 #[cfg(test)]
 mod tests {
    use super::{
@@ -0,0 +1,873 @@
+#![allow(dead_code)]
+use std::env;
+use std::fmt::{Display, Formatter};
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::time::UNIX_EPOCH;
+
+use crate::session::{Session, SessionError};
+
+/// Per-worktree session store that namespaces on-disk session files by
+/// workspace fingerprint so that parallel `opencode serve` instances never
+/// collide.
+///
+/// Create via [`SessionStore::from_cwd`] (derives the store path from the
+/// server's working directory) or [`SessionStore::from_data_dir`] (honours an
+/// explicit `--data-dir` flag).  Both constructors produce a directory layout
+/// of `<data_dir>/sessions/<workspace_hash>/` where `<workspace_hash>` is a
+/// stable hex digest of the canonical workspace root.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct SessionStore {
+    /// Resolved root of the session namespace, e.g.
+    /// `/home/user/project/.claw/sessions/a1b2c3d4e5f60718/`.
+    sessions_root: PathBuf,
+    /// The canonical workspace path that was fingerprinted.
+    workspace_root: PathBuf,
+}
+
+impl SessionStore {
+    /// Build a store from the server's current working directory.
+    ///
+    /// The on-disk layout becomes `<cwd>/.claw/sessions/<workspace_hash>/`.
+    pub fn from_cwd(cwd: impl AsRef<Path>) -> Result<Self, SessionControlError> {
+        let cwd = cwd.as_ref();
+        let sessions_root = cwd
+            .join(".claw")
+            .join("sessions")
+            .join(workspace_fingerprint(cwd));
+        fs::create_dir_all(&sessions_root)?;
+        Ok(Self {
+            sessions_root,
+            workspace_root: cwd.to_path_buf(),
+        })
+    }
+
+    /// Build a store from an explicit `--data-dir` flag.
+    ///
+    /// The on-disk layout becomes `<data_dir>/sessions/<workspace_hash>/`
+    /// where `<workspace_hash>` is derived from `workspace_root`.
+    pub fn from_data_dir(
+        data_dir: impl AsRef<Path>,
+        workspace_root: impl AsRef<Path>,
+    ) -> Result<Self, SessionControlError> {
+        let workspace_root = workspace_root.as_ref();
+        let sessions_root = data_dir
+            .as_ref()
+            .join("sessions")
+            .join(workspace_fingerprint(workspace_root));
+        fs::create_dir_all(&sessions_root)?;
+        Ok(Self {
+            sessions_root,
+            workspace_root: workspace_root.to_path_buf(),
+        })
+    }
+
+    /// The fully resolved sessions directory for this namespace.
+    #[must_use]
+    pub fn sessions_dir(&self) -> &Path {
+        &self.sessions_root
+    }
+
+    /// The workspace root this store is bound to.
+    #[must_use]
+    pub fn workspace_root(&self) -> &Path {
+        &self.workspace_root
+    }
+
+    pub fn create_handle(&self, session_id: &str) -> SessionHandle {
+        let id = session_id.to_string();
+        let path = self
+            .sessions_root
+            .join(format!("{id}.{PRIMARY_SESSION_EXTENSION}"));
+        SessionHandle { id, path }
+    }
+
+    pub fn resolve_reference(&self, reference: &str) -> Result<SessionHandle, SessionControlError> {
+        if is_session_reference_alias(reference) {
+            let latest = self.latest_session()?;
+            return Ok(SessionHandle {
+                id: latest.id,
+                path: latest.path,
+            });
+        }
+
+        let direct = PathBuf::from(reference);
+        let candidate = if direct.is_absolute() {
+            direct.clone()
+        } else {
+            self.workspace_root.join(&direct)
+        };
+        let looks_like_path = direct.extension().is_some() || direct.components().count() > 1;
+        let path = if candidate.exists() {
+            candidate
+        } else if looks_like_path {
+            return Err(SessionControlError::Format(
+                format_missing_session_reference(reference),
+            ));
+        } else {
+            self.resolve_managed_path(reference)?
+        };
+
+        Ok(SessionHandle {
+            id: session_id_from_path(&path).unwrap_or_else(|| reference.to_string()),
+            path,
+        })
+    }
+
+    pub fn resolve_managed_path(&self, session_id: &str) -> Result<PathBuf, SessionControlError> {
+        for extension in [PRIMARY_SESSION_EXTENSION, LEGACY_SESSION_EXTENSION] {
+            let path = self.sessions_root.join(format!("{session_id}.{extension}"));
+            if path.exists() {
+                return Ok(path);
+            }
+        }
+        Err(SessionControlError::Format(
+            format_missing_session_reference(session_id),
+        ))
+    }
+
+    pub fn list_sessions(&self) -> Result<Vec<ManagedSessionSummary>, SessionControlError> {
+        let mut sessions = Vec::new();
+        let read_result = fs::read_dir(&self.sessions_root);
+        let entries = match read_result {
+            Ok(entries) => entries,
+            Err(err) if err.kind() == std::io::ErrorKind::NotFound => return Ok(sessions),
+            Err(err) => return Err(err.into()),
+        };
+        for entry in entries {
+            let entry = entry?;
+            let path = entry.path();
+            if !is_managed_session_file(&path) {
+                continue;
+            }
+            let metadata = entry.metadata()?;
+            let modified_epoch_millis = metadata
+                .modified()
+                .ok()
+                .and_then(|time| time.duration_since(UNIX_EPOCH).ok())
+                .map(|duration| duration.as_millis())
+                .unwrap_or_default();
+            let (id, message_count, parent_session_id, branch_name) =
+                match Session::load_from_path(&path) {
+                    Ok(session) => {
+                        let parent_session_id = session
+                            .fork
+                            .as_ref()
+                            .map(|fork| fork.parent_session_id.clone());
+                        let branch_name = session
+                            .fork
+                            .as_ref()
+                            .and_then(|fork| fork.branch_name.clone());
+                        (
+                            session.session_id,
+                            session.messages.len(),
+                            parent_session_id,
+                            branch_name,
+                        )
+                    }
+                    Err(_) => (
+                        path.file_stem()
+                            .and_then(|value| value.to_str())
+                            .unwrap_or("unknown")
+                            .to_string(),
+                        0,
+                        None,
+                        None,
+                    ),
+                };
+            sessions.push(ManagedSessionSummary {
+                id,
+                path,
+                modified_epoch_millis,
+                message_count,
+                parent_session_id,
+                branch_name,
+            });
+        }
+        sessions.sort_by(|left, right| {
+            right
+                .modified_epoch_millis
+                .cmp(&left.modified_epoch_millis)
+                .then_with(|| right.id.cmp(&left.id))
+        });
+        Ok(sessions)
+    }
+
+    pub fn latest_session(&self) -> Result<ManagedSessionSummary, SessionControlError> {
+        self.list_sessions()?
+            .into_iter()
+            .next()
+            .ok_or_else(|| SessionControlError::Format(format_no_managed_sessions()))
+    }
+
+    pub fn load_session(
+        &self,
+        reference: &str,
+    ) -> Result<LoadedManagedSession, SessionControlError> {
+        let handle = self.resolve_reference(reference)?;
+        let session = Session::load_from_path(&handle.path)?;
+        Ok(LoadedManagedSession {
+            handle: SessionHandle {
+                id: session.session_id.clone(),
+                path: handle.path,
+            },
+            session,
+        })
+    }
+
+    pub fn fork_session(
+        &self,
+        session: &Session,
+        branch_name: Option<String>,
+    ) -> Result<ForkedManagedSession, SessionControlError> {
+        let parent_session_id = session.session_id.clone();
+        let forked = session.fork(branch_name);
+        let handle = self.create_handle(&forked.session_id);
+        let branch_name = forked
+            .fork
+            .as_ref()
+            .and_then(|fork| fork.branch_name.clone());
+        let forked = forked.with_persistence_path(handle.path.clone());
+        forked.save_to_path(&handle.path)?;
+        Ok(ForkedManagedSession {
+            parent_session_id,
+            handle,
+            session: forked,
+            branch_name,
+        })
+    }
+}
+
+/// Stable hex fingerprint of a workspace path.
+///
+/// Uses FNV-1a (64-bit) to produce a 16-char hex string that partitions the
+/// on-disk session directory per workspace root.
+#[must_use]
+pub fn workspace_fingerprint(workspace_root: &Path) -> String {
+    let input = workspace_root.to_string_lossy();
+    let mut hash = 0xcbf2_9ce4_8422_2325_u64;
+    for byte in input.as_bytes() {
+        hash ^= u64::from(*byte);
+        hash = hash.wrapping_mul(0x0100_0000_01b3);
+    }
+    format!("{hash:016x}")
+}
+
+pub const PRIMARY_SESSION_EXTENSION: &str = "jsonl";
+pub const LEGACY_SESSION_EXTENSION: &str = "json";
+pub const LATEST_SESSION_REFERENCE: &str = "latest";
+
+const SESSION_REFERENCE_ALIASES: &[&str] = &[LATEST_SESSION_REFERENCE, "last", "recent"];
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct SessionHandle {
+    pub id: String,
+    pub path: PathBuf,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ManagedSessionSummary {
+    pub id: String,
+    pub path: PathBuf,
+    pub modified_epoch_millis: u128,
+    pub message_count: usize,
+    pub parent_session_id: Option<String>,
+    pub branch_name: Option<String>,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct LoadedManagedSession {
+    pub handle: SessionHandle,
+    pub session: Session,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ForkedManagedSession {
+    pub parent_session_id: String,
+    pub handle: SessionHandle,
+    pub session: Session,
+    pub branch_name: Option<String>,
+}
+
+#[derive(Debug)]
+pub enum SessionControlError {
+    Io(std::io::Error),
+    Session(SessionError),
+    Format(String),
+}
+
+impl Display for SessionControlError {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Io(error) => write!(f, "{error}"),
+            Self::Session(error) => write!(f, "{error}"),
+            Self::Format(error) => write!(f, "{error}"),
+        }
+    }
+}
+
+impl std::error::Error for SessionControlError {}
+
+impl From<std::io::Error> for SessionControlError {
+    fn from(value: std::io::Error) -> Self {
+        Self::Io(value)
+    }
+}
+
+impl From<SessionError> for SessionControlError {
+    fn from(value: SessionError) -> Self {
+        Self::Session(value)
+    }
+}
+
+pub fn sessions_dir() -> Result<PathBuf, SessionControlError> {
+    managed_sessions_dir_for(env::current_dir()?)
+}
+
+pub fn managed_sessions_dir_for(
+    base_dir: impl AsRef<Path>,
+) -> Result<PathBuf, SessionControlError> {
+    let path = base_dir.as_ref().join(".claw").join("sessions");
+    fs::create_dir_all(&path)?;
+    Ok(path)
+}
+
+pub fn create_managed_session_handle(
+    session_id: &str,
+) -> Result<SessionHandle, SessionControlError> {
+    create_managed_session_handle_for(env::current_dir()?, session_id)
+}
+
+pub fn create_managed_session_handle_for(
+    base_dir: impl AsRef<Path>,
+    session_id: &str,
+) -> Result<SessionHandle, SessionControlError> {
+    let id = session_id.to_string();
+    let path =
+        managed_sessions_dir_for(base_dir)?.join(format!("{id}.{PRIMARY_SESSION_EXTENSION}"));
+    Ok(SessionHandle { id, path })
+}
+
+pub fn resolve_session_reference(reference: &str) -> Result<SessionHandle, SessionControlError> {
+    resolve_session_reference_for(env::current_dir()?, reference)
+}
+
+pub fn resolve_session_reference_for(
+    base_dir: impl AsRef<Path>,
+    reference: &str,
+) -> Result<SessionHandle, SessionControlError> {
+    let base_dir = base_dir.as_ref();
+    if is_session_reference_alias(reference) {
+        let latest = latest_managed_session_for(base_dir)?;
+        return Ok(SessionHandle {
+            id: latest.id,
+            path: latest.path,
+        });
+    }
+
+    let direct = PathBuf::from(reference);
+    let candidate = if direct.is_absolute() {
+        direct.clone()
+    } else {
+        base_dir.join(&direct)
+    };
+    let looks_like_path = direct.extension().is_some() || direct.components().count() > 1;
+    let path = if candidate.exists() {
+        candidate
+    } else if looks_like_path {
+        return Err(SessionControlError::Format(
+            format_missing_session_reference(reference),
+        ));
+    } else {
+        resolve_managed_session_path_for(base_dir, reference)?
+    };
+
+    Ok(SessionHandle {
+        id: session_id_from_path(&path).unwrap_or_else(|| reference.to_string()),
+        path,
+    })
+}
+
+pub fn resolve_managed_session_path(session_id: &str) -> Result<PathBuf, SessionControlError> {
+    resolve_managed_session_path_for(env::current_dir()?, session_id)
+}
+
+pub fn resolve_managed_session_path_for(
+    base_dir: impl AsRef<Path>,
+    session_id: &str,
+) -> Result<PathBuf, SessionControlError> {
+    let directory = managed_sessions_dir_for(base_dir)?;
+    for extension in [PRIMARY_SESSION_EXTENSION, LEGACY_SESSION_EXTENSION] {
+        let path = directory.join(format!("{session_id}.{extension}"));
+        if path.exists() {
+            return Ok(path);
+        }
+    }
+    Err(SessionControlError::Format(
+        format_missing_session_reference(session_id),
+    ))
+}
+
+#[must_use]
+pub fn is_managed_session_file(path: &Path) -> bool {
+    path.extension()
+        .and_then(|ext| ext.to_str())
+        .is_some_and(|extension| {
+            extension == PRIMARY_SESSION_EXTENSION || extension == LEGACY_SESSION_EXTENSION
+        })
+}
+
+pub fn list_managed_sessions() -> Result<Vec<ManagedSessionSummary>, SessionControlError> {
+    list_managed_sessions_for(env::current_dir()?)
+}
+
+pub fn list_managed_sessions_for(
+    base_dir: impl AsRef<Path>,
+) -> Result<Vec<ManagedSessionSummary>, SessionControlError> {
+    let mut sessions = Vec::new();
+    for entry in fs::read_dir(managed_sessions_dir_for(base_dir)?)? {
+        let entry = entry?;
+        let path = entry.path();
+        if !is_managed_session_file(&path) {
+            continue;
+        }
+        let metadata = entry.metadata()?;
+        let modified_epoch_millis = metadata
+            .modified()
+            .ok()
+            .and_then(|time| time.duration_since(UNIX_EPOCH).ok())
+            .map(|duration| duration.as_millis())
+            .unwrap_or_default();
+        let (id, message_count, parent_session_id, branch_name) =
+            match Session::load_from_path(&path) {
+                Ok(session) => {
+                    let parent_session_id = session
+                        .fork
+                        .as_ref()
+                        .map(|fork| fork.parent_session_id.clone());
+                    let branch_name = session
+                        .fork
+                        .as_ref()
+                        .and_then(|fork| fork.branch_name.clone());
+                    (
+                        session.session_id,
+                        session.messages.len(),
+                        parent_session_id,
+                        branch_name,
+                    )
+                }
+                Err(_) => (
+                    path.file_stem()
+                        .and_then(|value| value.to_str())
+                        .unwrap_or("unknown")
+                        .to_string(),
+                    0,
+                    None,
+                    None,
+                ),
+            };
+        sessions.push(ManagedSessionSummary {
+            id,
+            path,
+            modified_epoch_millis,
+            message_count,
+            parent_session_id,
+            branch_name,
+        });
+    }
+    sessions.sort_by(|left, right| {
+        right
+            .modified_epoch_millis
+            .cmp(&left.modified_epoch_millis)
+            .then_with(|| right.id.cmp(&left.id))
+    });
+    Ok(sessions)
+}
+
+pub fn latest_managed_session() -> Result<ManagedSessionSummary, SessionControlError> {
+    latest_managed_session_for(env::current_dir()?)
+}
+
+pub fn latest_managed_session_for(
+    base_dir: impl AsRef<Path>,
+) -> Result<ManagedSessionSummary, SessionControlError> {
+    list_managed_sessions_for(base_dir)?
+        .into_iter()
+        .next()
+        .ok_or_else(|| SessionControlError::Format(format_no_managed_sessions()))
+}
+
+pub fn load_managed_session(reference: &str) -> Result<LoadedManagedSession, SessionControlError> {
+    load_managed_session_for(env::current_dir()?, reference)
+}
+
+pub fn load_managed_session_for(
+    base_dir: impl AsRef<Path>,
+    reference: &str,
+) -> Result<LoadedManagedSession, SessionControlError> {
+    let handle = resolve_session_reference_for(base_dir, reference)?;
+    let session = Session::load_from_path(&handle.path)?;
+    Ok(LoadedManagedSession {
+        handle: SessionHandle {
+            id: session.session_id.clone(),
+            path: handle.path,
+        },
+        session,
+    })
+}
+
+pub fn fork_managed_session(
+    session: &Session,
+    branch_name: Option<String>,
+) -> Result<ForkedManagedSession, SessionControlError> {
+    fork_managed_session_for(env::current_dir()?, session, branch_name)
+}
+
+pub fn fork_managed_session_for(
+    base_dir: impl AsRef<Path>,
+    session: &Session,
+    branch_name: Option<String>,
+) -> Result<ForkedManagedSession, SessionControlError> {
+    let parent_session_id = session.session_id.clone();
+    let forked = session.fork(branch_name);
+    let handle = create_managed_session_handle_for(base_dir, &forked.session_id)?;
+    let branch_name = forked
+        .fork
+        .as_ref()
+        .and_then(|fork| fork.branch_name.clone());
+    let forked = forked.with_persistence_path(handle.path.clone());
+    forked.save_to_path(&handle.path)?;
+    Ok(ForkedManagedSession {
+        parent_session_id,
+        handle,
+        session: forked,
+        branch_name,
+    })
+}
+
+#[must_use]
+pub fn is_session_reference_alias(reference: &str) -> bool {
+    SESSION_REFERENCE_ALIASES
+        .iter()
+        .any(|alias| reference.eq_ignore_ascii_case(alias))
+}
+
+fn session_id_from_path(path: &Path) -> Option<String> {
+    path.file_name()
+        .and_then(|value| value.to_str())
+        .and_then(|name| {
+            name.strip_suffix(&format!(".{PRIMARY_SESSION_EXTENSION}"))
+                .or_else(|| name.strip_suffix(&format!(".{LEGACY_SESSION_EXTENSION}")))
+        })
+        .map(ToOwned::to_owned)
+}
+
+fn format_missing_session_reference(reference: &str) -> String {
+    format!(
+        "session not found: {reference}\nHint: managed sessions live in .claw/sessions/. Try `{LATEST_SESSION_REFERENCE}` for the most recent session or `/session list` in the REPL."
+    )
+}
+
+fn format_no_managed_sessions() -> String {
+    format!(
+        "no managed sessions found in .claw/sessions/\nStart `claw` to create a session, then rerun with `--resume {LATEST_SESSION_REFERENCE}`."
+    )
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{
+        create_managed_session_handle_for, fork_managed_session_for, is_session_reference_alias,
+        list_managed_sessions_for, load_managed_session_for, resolve_session_reference_for,
+        workspace_fingerprint, ManagedSessionSummary, SessionStore, LATEST_SESSION_REFERENCE,
+    };
+    use crate::session::Session;
+    use std::fs;
+    use std::path::{Path, PathBuf};
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn temp_dir() -> PathBuf {
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_nanos();
+        std::env::temp_dir().join(format!("runtime-session-control-{nanos}"))
+    }
+
+    fn persist_session(root: &Path, text: &str) -> Session {
+        let mut session = Session::new();
+        session
+            .push_user_text(text)
+            .expect("session message should save");
+        let handle = create_managed_session_handle_for(root, &session.session_id)
+            .expect("managed session handle should build");
+        let session = session.with_persistence_path(handle.path.clone());
+        session
+            .save_to_path(&handle.path)
+            .expect("session should persist");
+        session
+    }
+
+    fn wait_for_next_millisecond() {
+        let start = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_millis();
+        while SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_millis()
+            <= start
+        {}
+    }
+
+    fn summary_by_id<'a>(
+        summaries: &'a [ManagedSessionSummary],
+        id: &str,
+    ) -> &'a ManagedSessionSummary {
+        summaries
+            .iter()
+            .find(|summary| summary.id == id)
+            .expect("session summary should exist")
+    }
+
+    #[test]
+    fn creates_and_lists_managed_sessions() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("root dir should exist");
+        let older = persist_session(&root, "older session");
+        wait_for_next_millisecond();
+        let newer = persist_session(&root, "newer session");
+
+        // when
+        let sessions = list_managed_sessions_for(&root).expect("managed sessions should list");
+
+        // then
+        assert_eq!(sessions.len(), 2);
+        assert_eq!(sessions[0].id, newer.session_id);
+        assert_eq!(summary_by_id(&sessions, &older.session_id).message_count, 1);
+        assert_eq!(summary_by_id(&sessions, &newer.session_id).message_count, 1);
+        fs::remove_dir_all(root).expect("temp dir should clean up");
+    }
+
+    #[test]
+    fn resolves_latest_alias_and_loads_session_from_workspace_root() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("root dir should exist");
+        let older = persist_session(&root, "older session");
+        wait_for_next_millisecond();
+        let newer = persist_session(&root, "newer session");
+
+        // when
+        let handle = resolve_session_reference_for(&root, LATEST_SESSION_REFERENCE)
+            .expect("latest alias should resolve");
+        let loaded = load_managed_session_for(&root, "recent")
+            .expect("recent alias should load the latest session");
+
+        // then
+        assert_eq!(handle.id, newer.session_id);
+        assert_eq!(loaded.handle.id, newer.session_id);
+        assert_eq!(loaded.session.messages.len(), 1);
+        assert_ne!(loaded.handle.id, older.session_id);
+        assert!(is_session_reference_alias("last"));
+        fs::remove_dir_all(root).expect("temp dir should clean up");
+    }
+
+    #[test]
+    fn forks_session_into_managed_storage_with_lineage() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("root dir should exist");
+        let source = persist_session(&root, "parent session");
+
+        // when
+        let forked = fork_managed_session_for(&root, &source, Some("incident-review".to_string()))
+            .expect("session should fork");
+        let sessions = list_managed_sessions_for(&root).expect("managed sessions should list");
+        let summary = summary_by_id(&sessions, &forked.handle.id);
+
+        // then
+        assert_eq!(forked.parent_session_id, source.session_id);
+        assert_eq!(forked.branch_name.as_deref(), Some("incident-review"));
+        assert_eq!(
+            summary.parent_session_id.as_deref(),
+            Some(source.session_id.as_str())
+        );
+        assert_eq!(summary.branch_name.as_deref(), Some("incident-review"));
+        assert_eq!(
+            forked.session.persistence_path(),
+            Some(forked.handle.path.as_path())
+        );
+        fs::remove_dir_all(root).expect("temp dir should clean up");
+    }
+
+    // ------------------------------------------------------------------
+    // Per-worktree session isolation (SessionStore) tests
+    // ------------------------------------------------------------------
+
+    fn persist_session_via_store(store: &SessionStore, text: &str) -> Session {
+        let mut session = Session::new();
+        session
+            .push_user_text(text)
+            .expect("session message should save");
+        let handle = store.create_handle(&session.session_id);
+        let session = session.with_persistence_path(handle.path.clone());
+        session
+            .save_to_path(&handle.path)
+            .expect("session should persist");
+        session
+    }
+
+    #[test]
+    fn workspace_fingerprint_is_deterministic_and_differs_per_path() {
+        // given
+        let path_a = Path::new("/tmp/worktree-alpha");
+        let path_b = Path::new("/tmp/worktree-beta");
+
+        // when
+        let fp_a1 = workspace_fingerprint(path_a);
+        let fp_a2 = workspace_fingerprint(path_a);
+        let fp_b = workspace_fingerprint(path_b);
+
+        // then
+        assert_eq!(fp_a1, fp_a2, "same path must produce the same fingerprint");
+        assert_ne!(
+            fp_a1, fp_b,
+            "different paths must produce different fingerprints"
+        );
+        assert_eq!(fp_a1.len(), 16, "fingerprint must be a 16-char hex string");
+    }
+
+    #[test]
+    fn session_store_from_cwd_isolates_sessions_by_workspace() {
+        // given
+        let base = temp_dir();
+        let workspace_a = base.join("repo-alpha");
+        let workspace_b = base.join("repo-beta");
+        fs::create_dir_all(&workspace_a).expect("workspace a should exist");
+        fs::create_dir_all(&workspace_b).expect("workspace b should exist");
+
+        let store_a = SessionStore::from_cwd(&workspace_a).expect("store a should build");
+        let store_b = SessionStore::from_cwd(&workspace_b).expect("store b should build");
+
+        // when
+        let session_a = persist_session_via_store(&store_a, "alpha work");
+        let _session_b = persist_session_via_store(&store_b, "beta work");
+
+        // then — each store only sees its own sessions
+        let list_a = store_a.list_sessions().expect("list a");
+        let list_b = store_b.list_sessions().expect("list b");
+        assert_eq!(list_a.len(), 1, "store a should see exactly one session");
+        assert_eq!(list_b.len(), 1, "store b should see exactly one session");
+        assert_eq!(list_a[0].id, session_a.session_id);
+        assert_ne!(
+            store_a.sessions_dir(),
+            store_b.sessions_dir(),
+            "session directories must differ across workspaces"
+        );
+        fs::remove_dir_all(base).expect("temp dir should clean up");
+    }
+
+    #[test]
+    fn session_store_from_data_dir_namespaces_by_workspace() {
+        // given
+        let base = temp_dir();
+        let data_dir = base.join("global-data");
+        let workspace_a = PathBuf::from("/tmp/project-one");
+        let workspace_b = PathBuf::from("/tmp/project-two");
+        fs::create_dir_all(&data_dir).expect("data dir should exist");
+
+        let store_a =
+            SessionStore::from_data_dir(&data_dir, &workspace_a).expect("store a should build");
+        let store_b =
+            SessionStore::from_data_dir(&data_dir, &workspace_b).expect("store b should build");
+
+        // when
+        persist_session_via_store(&store_a, "work in project-one");
+        persist_session_via_store(&store_b, "work in project-two");
+
+        // then
+        assert_ne!(
+            store_a.sessions_dir(),
+            store_b.sessions_dir(),
+            "data-dir stores must namespace by workspace"
+        );
+        assert_eq!(store_a.list_sessions().expect("list a").len(), 1);
+        assert_eq!(store_b.list_sessions().expect("list b").len(), 1);
+        assert_eq!(store_a.workspace_root(), workspace_a.as_path());
+        assert_eq!(store_b.workspace_root(), workspace_b.as_path());
+        fs::remove_dir_all(base).expect("temp dir should clean up");
+    }
+
+    #[test]
+    fn session_store_create_and_load_round_trip() {
+        // given
+        let base = temp_dir();
+        fs::create_dir_all(&base).expect("base dir should exist");
+        let store = SessionStore::from_cwd(&base).expect("store should build");
+        let session = persist_session_via_store(&store, "round-trip message");
+
+        // when
+        let loaded = store
+            .load_session(&session.session_id)
+            .expect("session should load via store");
+
+        // then
+        assert_eq!(loaded.handle.id, session.session_id);
+        assert_eq!(loaded.session.messages.len(), 1);
+        fs::remove_dir_all(base).expect("temp dir should clean up");
+    }
+
+    #[test]
+    fn session_store_latest_and_resolve_reference() {
+        // given
+        let base = temp_dir();
+        fs::create_dir_all(&base).expect("base dir should exist");
+        let store = SessionStore::from_cwd(&base).expect("store should build");
+        let _older = persist_session_via_store(&store, "older");
+        wait_for_next_millisecond();
+        let newer = persist_session_via_store(&store, "newer");
+
+        // when
+        let latest = store.latest_session().expect("latest should resolve");
+        let handle = store
+            .resolve_reference("latest")
+            .expect("latest alias should resolve");
+
+        // then
+        assert_eq!(latest.id, newer.session_id);
+        assert_eq!(handle.id, newer.session_id);
+        fs::remove_dir_all(base).expect("temp dir should clean up");
+    }
+
+    #[test]
+    fn session_store_fork_stays_in_same_namespace() {
+        // given
+        let base = temp_dir();
+        fs::create_dir_all(&base).expect("base dir should exist");
+        let store = SessionStore::from_cwd(&base).expect("store should build");
+        let source = persist_session_via_store(&store, "parent work");
+
+        // when
+        let forked = store
+            .fork_session(&source, Some("bugfix".to_string()))
+            .expect("fork should succeed");
+        let sessions = store.list_sessions().expect("list sessions");
+
+        // then
+        assert_eq!(
+            sessions.len(),
+            2,
+            "forked session must land in the same namespace"
+        );
+        assert_eq!(forked.parent_session_id, source.session_id);
+        assert_eq!(forked.branch_name.as_deref(), Some("bugfix"));
+        assert!(
+            forked.handle.path.starts_with(store.sessions_dir()),
+            "forked session path must be inside the store namespace"
+        );
+        fs::remove_dir_all(base).expect("temp dir should clean up");
+    }
+}
@@ -80,7 +80,11 @@ impl IncrementalSseParser {
    }

    fn take_event(&mut self) -> Option<SseEvent> {
-        if self.data_lines.is_empty() && self.event_name.is_none() && self.id.is_none() && self.retry.is_none() {
+        if self.data_lines.is_empty()
+            && self.event_name.is_none()
+            && self.id.is_none()
+            && self.retry.is_none()
+        {
            return None;
        }

@@ -102,8 +106,13 @@ mod tests {

    #[test]
    fn parses_streaming_events() {
+        // given
        let mut parser = IncrementalSseParser::new();
+
+        // when
        let first = parser.push_chunk("event: message\ndata: hel");
+
+        // then
        assert!(first.is_empty());

        let second = parser.push_chunk("lo\n\nid: 1\ndata: world\n\n");
@@ -125,4 +134,25 @@ mod tests {
            ]
        );
    }
+
+    #[test]
+    fn finish_flushes_a_trailing_event_without_separator() {
+        // given
+        let mut parser = IncrementalSseParser::new();
+        parser.push_chunk("event: message\ndata: trailing");
+
+        // when
+        let events = parser.finish();
+
+        // then
+        assert_eq!(
+            events,
+            vec![SseEvent {
+                event: Some("message".to_string()),
+                data: "trailing".to_string(),
+                id: None,
+                retry: None,
+            }]
+        );
+    }
 }
@@ -0,0 +1,429 @@
+#![allow(clippy::must_use_candidate)]
+use std::path::Path;
+use std::process::Command;
+
+/// Outcome of comparing the worktree HEAD against the expected base commit.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum BaseCommitState {
+    /// HEAD matches the expected base commit.
+    Matches,
+    /// HEAD has diverged from the expected base.
+    Diverged { expected: String, actual: String },
+    /// No expected base was supplied (neither flag nor file).
+    NoExpectedBase,
+    /// The working directory is not inside a git repository.
+    NotAGitRepo,
+}
+
+/// Where the expected base commit originated from.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum BaseCommitSource {
+    Flag(String),
+    File(String),
+}
+
+/// Read the `.claw-base` file from the given directory and return the trimmed
+/// commit hash, or `None` when the file is absent or empty.
+pub fn read_claw_base_file(cwd: &Path) -> Option<String> {
+    let path = cwd.join(".claw-base");
+    let content = std::fs::read_to_string(path).ok()?;
+    let trimmed = content.trim();
+    if trimmed.is_empty() {
+        None
+    } else {
+        Some(trimmed.to_string())
+    }
+}
+
+/// Resolve the expected base commit: prefer the `--base-commit` flag value,
+/// fall back to reading `.claw-base` from `cwd`.
+pub fn resolve_expected_base(flag_value: Option<&str>, cwd: &Path) -> Option<BaseCommitSource> {
+    if let Some(value) = flag_value {
+        let trimmed = value.trim();
+        if !trimmed.is_empty() {
+            return Some(BaseCommitSource::Flag(trimmed.to_string()));
+        }
+    }
+    read_claw_base_file(cwd).map(BaseCommitSource::File)
+}
+
+/// Verify that the worktree HEAD matches `expected_base`.
+///
+/// Returns [`BaseCommitState::NoExpectedBase`] when no expected commit is
+/// provided (the check is effectively a no-op in that case).
+pub fn check_base_commit(cwd: &Path, expected_base: Option<&BaseCommitSource>) -> BaseCommitState {
+    let Some(source) = expected_base else {
+        return BaseCommitState::NoExpectedBase;
+    };
+    let expected_raw = match source {
+        BaseCommitSource::Flag(value) | BaseCommitSource::File(value) => value.as_str(),
+    };
+
+    let Some(head_sha) = resolve_head_sha(cwd) else {
+        return BaseCommitState::NotAGitRepo;
+    };
+
+    let Some(expected_sha) = resolve_rev(cwd, expected_raw) else {
+        // If the expected ref cannot be resolved, compare raw strings as a
+        // best-effort fallback (e.g. partial SHA provided by the caller).
+        return if head_sha.starts_with(expected_raw) || expected_raw.starts_with(&head_sha) {
+            BaseCommitState::Matches
+        } else {
+            BaseCommitState::Diverged {
+                expected: expected_raw.to_string(),
+                actual: head_sha,
+            }
+        };
+    };
+
+    if head_sha == expected_sha {
+        BaseCommitState::Matches
+    } else {
+        BaseCommitState::Diverged {
+            expected: expected_sha,
+            actual: head_sha,
+        }
+    }
+}
+
+/// Format a human-readable warning when the base commit has diverged.
+///
+/// Returns `None` for non-warning states (`Matches`, `NoExpectedBase`).
+pub fn format_stale_base_warning(state: &BaseCommitState) -> Option<String> {
+    match state {
+        BaseCommitState::Diverged { expected, actual } => Some(format!(
+            "warning: worktree HEAD ({actual}) does not match expected base commit ({expected}). \
+             Session may run against a stale codebase."
+        )),
+        BaseCommitState::NotAGitRepo => {
+            Some("warning: stale-base check skipped — not inside a git repository.".to_string())
+        }
+        BaseCommitState::Matches | BaseCommitState::NoExpectedBase => None,
+    }
+}
+
+fn resolve_head_sha(cwd: &Path) -> Option<String> {
+    resolve_rev(cwd, "HEAD")
+}
+
+fn resolve_rev(cwd: &Path, rev: &str) -> Option<String> {
+    let output = Command::new("git")
+        .args(["rev-parse", rev])
+        .current_dir(cwd)
+        .output()
+        .ok()?;
+    if !output.status.success() {
+        return None;
+    }
+    let sha = String::from_utf8(output.stdout).ok()?;
+    let trimmed = sha.trim();
+    if trimmed.is_empty() {
+        None
+    } else {
+        Some(trimmed.to_string())
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::fs;
+    use std::process::Command;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn temp_dir() -> std::path::PathBuf {
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_nanos();
+        std::env::temp_dir().join(format!("runtime-stale-base-{nanos}"))
+    }
+
+    fn init_repo(path: &std::path::Path) {
+        fs::create_dir_all(path).expect("create repo dir");
+        run(path, &["init", "--quiet", "-b", "main"]);
+        run(path, &["config", "user.email", "tests@example.com"]);
+        run(path, &["config", "user.name", "Stale Base Tests"]);
+        fs::write(path.join("init.txt"), "initial\n").expect("write init file");
+        run(path, &["add", "."]);
+        run(path, &["commit", "-m", "initial commit", "--quiet"]);
+    }
+
+    fn run(cwd: &std::path::Path, args: &[&str]) {
+        let status = Command::new("git")
+            .args(args)
+            .current_dir(cwd)
+            .status()
+            .unwrap_or_else(|e| panic!("git {} failed to execute: {e}", args.join(" ")));
+        assert!(
+            status.success(),
+            "git {} exited with {status}",
+            args.join(" ")
+        );
+    }
+
+    fn commit_file(repo: &std::path::Path, name: &str, msg: &str) {
+        fs::write(repo.join(name), format!("{msg}\n")).expect("write file");
+        run(repo, &["add", name]);
+        run(repo, &["commit", "-m", msg, "--quiet"]);
+    }
+
+    fn head_sha(repo: &std::path::Path) -> String {
+        let output = Command::new("git")
+            .args(["rev-parse", "HEAD"])
+            .current_dir(repo)
+            .output()
+            .expect("git rev-parse HEAD");
+        String::from_utf8(output.stdout)
+            .expect("valid utf8")
+            .trim()
+            .to_string()
+    }
+
+    #[test]
+    fn matches_when_head_equals_expected_base() {
+        // given
+        let root = temp_dir();
+        init_repo(&root);
+        let sha = head_sha(&root);
+        let source = BaseCommitSource::Flag(sha);
+
+        // when
+        let state = check_base_commit(&root, Some(&source));
+
+        // then
+        assert_eq!(state, BaseCommitState::Matches);
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn diverged_when_head_moved_past_expected_base() {
+        // given
+        let root = temp_dir();
+        init_repo(&root);
+        let old_sha = head_sha(&root);
+        commit_file(&root, "extra.txt", "move head forward");
+        let new_sha = head_sha(&root);
+        let source = BaseCommitSource::Flag(old_sha.clone());
+
+        // when
+        let state = check_base_commit(&root, Some(&source));
+
+        // then
+        assert_eq!(
+            state,
+            BaseCommitState::Diverged {
+                expected: old_sha,
+                actual: new_sha,
+            }
+        );
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn no_expected_base_when_source_is_none() {
+        // given
+        let root = temp_dir();
+        init_repo(&root);
+
+        // when
+        let state = check_base_commit(&root, None);
+
+        // then
+        assert_eq!(state, BaseCommitState::NoExpectedBase);
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn not_a_git_repo_when_outside_repo() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("create dir");
+        let source = BaseCommitSource::Flag("abc1234".to_string());
+
+        // when
+        let state = check_base_commit(&root, Some(&source));
+
+        // then
+        assert_eq!(state, BaseCommitState::NotAGitRepo);
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn reads_claw_base_file() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("create dir");
+        fs::write(root.join(".claw-base"), "abc1234def5678\n").expect("write .claw-base");
+
+        // when
+        let value = read_claw_base_file(&root);
+
+        // then
+        assert_eq!(value, Some("abc1234def5678".to_string()));
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn returns_none_for_missing_claw_base_file() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("create dir");
+
+        // when
+        let value = read_claw_base_file(&root);
+
+        // then
+        assert!(value.is_none());
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn returns_none_for_empty_claw_base_file() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("create dir");
+        fs::write(root.join(".claw-base"), "  \n").expect("write empty .claw-base");
+
+        // when
+        let value = read_claw_base_file(&root);
+
+        // then
+        assert!(value.is_none());
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn resolve_expected_base_prefers_flag_over_file() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("create dir");
+        fs::write(root.join(".claw-base"), "from_file\n").expect("write .claw-base");
+
+        // when
+        let source = resolve_expected_base(Some("from_flag"), &root);
+
+        // then
+        assert_eq!(
+            source,
+            Some(BaseCommitSource::Flag("from_flag".to_string()))
+        );
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn resolve_expected_base_falls_back_to_file() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("create dir");
+        fs::write(root.join(".claw-base"), "from_file\n").expect("write .claw-base");
+
+        // when
+        let source = resolve_expected_base(None, &root);
+
+        // then
+        assert_eq!(
+            source,
+            Some(BaseCommitSource::File("from_file".to_string()))
+        );
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn resolve_expected_base_returns_none_when_nothing_available() {
+        // given
+        let root = temp_dir();
+        fs::create_dir_all(&root).expect("create dir");
+
+        // when
+        let source = resolve_expected_base(None, &root);
+
+        // then
+        assert!(source.is_none());
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn format_warning_returns_message_for_diverged() {
+        // given
+        let state = BaseCommitState::Diverged {
+            expected: "abc1234".to_string(),
+            actual: "def5678".to_string(),
+        };
+
+        // when
+        let warning = format_stale_base_warning(&state);
+
+        // then
+        let message = warning.expect("should produce warning");
+        assert!(message.contains("abc1234"));
+        assert!(message.contains("def5678"));
+        assert!(message.contains("stale codebase"));
+    }
+
+    #[test]
+    fn format_warning_returns_none_for_matches() {
+        // given
+        let state = BaseCommitState::Matches;
+
+        // when
+        let warning = format_stale_base_warning(&state);
+
+        // then
+        assert!(warning.is_none());
+    }
+
+    #[test]
+    fn format_warning_returns_none_for_no_expected_base() {
+        // given
+        let state = BaseCommitState::NoExpectedBase;
+
+        // when
+        let warning = format_stale_base_warning(&state);
+
+        // then
+        assert!(warning.is_none());
+    }
+
+    #[test]
+    fn matches_with_claw_base_file_in_real_repo() {
+        // given
+        let root = temp_dir();
+        init_repo(&root);
+        let sha = head_sha(&root);
+        fs::write(root.join(".claw-base"), format!("{sha}\n")).expect("write .claw-base");
+        let source = resolve_expected_base(None, &root);
+
+        // when
+        let state = check_base_commit(&root, source.as_ref());
+
+        // then
+        assert_eq!(state, BaseCommitState::Matches);
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn diverged_with_claw_base_file_after_new_commit() {
+        // given
+        let root = temp_dir();
+        init_repo(&root);
+        let old_sha = head_sha(&root);
+        fs::write(root.join(".claw-base"), format!("{old_sha}\n")).expect("write .claw-base");
+        commit_file(&root, "new.txt", "advance head");
+        let new_sha = head_sha(&root);
+        let source = resolve_expected_base(None, &root);
+
+        // when
+        let state = check_base_commit(&root, source.as_ref());
+
+        // then
+        assert_eq!(
+            state,
+            BaseCommitState::Diverged {
+                expected: old_sha,
+                actual: new_sha,
+            }
+        );
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+}
@@ -0,0 +1,417 @@
+#![allow(clippy::must_use_candidate)]
+use std::path::Path;
+use std::process::Command;
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum BranchFreshness {
+    Fresh,
+    Stale {
+        commits_behind: usize,
+        missing_fixes: Vec<String>,
+    },
+    Diverged {
+        ahead: usize,
+        behind: usize,
+        missing_fixes: Vec<String>,
+    },
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum StaleBranchPolicy {
+    AutoRebase,
+    AutoMergeForward,
+    WarnOnly,
+    Block,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum StaleBranchEvent {
+    BranchStaleAgainstMain {
+        branch: String,
+        commits_behind: usize,
+        missing_fixes: Vec<String>,
+    },
+    RebaseAttempted {
+        branch: String,
+        result: String,
+    },
+    MergeForwardAttempted {
+        branch: String,
+        result: String,
+    },
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum StaleBranchAction {
+    Noop,
+    Warn { message: String },
+    Block { message: String },
+    Rebase,
+    MergeForward,
+}
+
+pub fn check_freshness(branch: &str, main_ref: &str) -> BranchFreshness {
+    check_freshness_in(branch, main_ref, Path::new("."))
+}
+
+pub fn apply_policy(freshness: &BranchFreshness, policy: StaleBranchPolicy) -> StaleBranchAction {
+    match freshness {
+        BranchFreshness::Fresh => StaleBranchAction::Noop,
+        BranchFreshness::Stale {
+            commits_behind,
+            missing_fixes,
+        } => match policy {
+            StaleBranchPolicy::WarnOnly => StaleBranchAction::Warn {
+                message: format!(
+                    "Branch is {commits_behind} commit(s) behind main. Missing fixes: {}",
+                    if missing_fixes.is_empty() {
+                        "(none)".to_string()
+                    } else {
+                        missing_fixes.join("; ")
+                    }
+                ),
+            },
+            StaleBranchPolicy::Block => StaleBranchAction::Block {
+                message: format!(
+                    "Branch is {commits_behind} commit(s) behind main and must be updated before proceeding."
+                ),
+            },
+            StaleBranchPolicy::AutoRebase => StaleBranchAction::Rebase,
+            StaleBranchPolicy::AutoMergeForward => StaleBranchAction::MergeForward,
+        },
+        BranchFreshness::Diverged {
+            ahead,
+            behind,
+            missing_fixes,
+        } => match policy {
+            StaleBranchPolicy::WarnOnly => StaleBranchAction::Warn {
+                message: format!(
+                    "Branch has diverged: {ahead} commit(s) ahead, {behind} commit(s) behind main. Missing fixes: {}",
+                    format_missing_fixes(missing_fixes)
+                ),
+            },
+            StaleBranchPolicy::Block => StaleBranchAction::Block {
+                message: format!(
+                    "Branch has diverged ({ahead} ahead, {behind} behind) and must be reconciled before proceeding. Missing fixes: {}",
+                    format_missing_fixes(missing_fixes)
+                ),
+            },
+            StaleBranchPolicy::AutoRebase => StaleBranchAction::Rebase,
+            StaleBranchPolicy::AutoMergeForward => StaleBranchAction::MergeForward,
+        },
+    }
+}
+
+pub(crate) fn check_freshness_in(
+    branch: &str,
+    main_ref: &str,
+    repo_path: &Path,
+) -> BranchFreshness {
+    let behind = rev_list_count(main_ref, branch, repo_path);
+    let ahead = rev_list_count(branch, main_ref, repo_path);
+
+    if behind == 0 {
+        return BranchFreshness::Fresh;
+    }
+
+    if ahead > 0 {
+        return BranchFreshness::Diverged {
+            ahead,
+            behind,
+            missing_fixes: missing_fix_subjects(main_ref, branch, repo_path),
+        };
+    }
+
+    let missing_fixes = missing_fix_subjects(main_ref, branch, repo_path);
+    BranchFreshness::Stale {
+        commits_behind: behind,
+        missing_fixes,
+    }
+}
+
+fn format_missing_fixes(missing_fixes: &[String]) -> String {
+    if missing_fixes.is_empty() {
+        "(none)".to_string()
+    } else {
+        missing_fixes.join("; ")
+    }
+}
+
+fn rev_list_count(a: &str, b: &str, repo_path: &Path) -> usize {
+    let output = Command::new("git")
+        .args(["rev-list", "--count", &format!("{b}..{a}")])
+        .current_dir(repo_path)
+        .output();
+    match output {
+        Ok(o) if o.status.success() => String::from_utf8_lossy(&o.stdout)
+            .trim()
+            .parse::<usize>()
+            .unwrap_or(0),
+        _ => 0,
+    }
+}
+
+fn missing_fix_subjects(a: &str, b: &str, repo_path: &Path) -> Vec<String> {
+    let output = Command::new("git")
+        .args(["log", "--format=%s", &format!("{b}..{a}")])
+        .current_dir(repo_path)
+        .output();
+    match output {
+        Ok(o) if o.status.success() => String::from_utf8_lossy(&o.stdout)
+            .lines()
+            .filter(|l| !l.is_empty())
+            .map(String::from)
+            .collect(),
+        _ => Vec::new(),
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::fs;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn temp_dir() -> std::path::PathBuf {
+        let nanos = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .expect("time should be after epoch")
+            .as_nanos();
+        std::env::temp_dir().join(format!("runtime-stale-branch-{nanos}"))
+    }
+
+    fn init_repo(path: &Path) {
+        fs::create_dir_all(path).expect("create repo dir");
+        run(path, &["init", "--quiet", "-b", "main"]);
+        run(path, &["config", "user.email", "tests@example.com"]);
+        run(path, &["config", "user.name", "Stale Branch Tests"]);
+        fs::write(path.join("init.txt"), "initial\n").expect("write init file");
+        run(path, &["add", "."]);
+        run(path, &["commit", "-m", "initial commit", "--quiet"]);
+    }
+
+    fn run(cwd: &Path, args: &[&str]) {
+        let status = Command::new("git")
+            .args(args)
+            .current_dir(cwd)
+            .status()
+            .unwrap_or_else(|e| panic!("git {} failed to execute: {e}", args.join(" ")));
+        assert!(
+            status.success(),
+            "git {} exited with {status}",
+            args.join(" ")
+        );
+    }
+
+    fn commit_file(repo: &Path, name: &str, msg: &str) {
+        fs::write(repo.join(name), format!("{msg}\n")).expect("write file");
+        run(repo, &["add", name]);
+        run(repo, &["commit", "-m", msg, "--quiet"]);
+    }
+
+    #[test]
+    fn fresh_branch_passes() {
+        let root = temp_dir();
+        init_repo(&root);
+
+        // given
+        run(&root, &["checkout", "-b", "topic"]);
+
+        // when
+        let freshness = check_freshness_in("topic", "main", &root);
+
+        // then
+        assert_eq!(freshness, BranchFreshness::Fresh);
+
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn fresh_branch_ahead_of_main_still_fresh() {
+        let root = temp_dir();
+        init_repo(&root);
+
+        // given
+        run(&root, &["checkout", "-b", "topic"]);
+        commit_file(&root, "feature.txt", "add feature");
+
+        // when
+        let freshness = check_freshness_in("topic", "main", &root);
+
+        // then
+        assert_eq!(freshness, BranchFreshness::Fresh);
+
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn stale_branch_detected_with_correct_behind_count_and_missing_fixes() {
+        let root = temp_dir();
+        init_repo(&root);
+
+        // given
+        run(&root, &["checkout", "-b", "topic"]);
+        run(&root, &["checkout", "main"]);
+        commit_file(&root, "fix1.txt", "fix: resolve timeout");
+        commit_file(&root, "fix2.txt", "fix: handle null pointer");
+
+        // when
+        let freshness = check_freshness_in("topic", "main", &root);
+
+        // then
+        match freshness {
+            BranchFreshness::Stale {
+                commits_behind,
+                missing_fixes,
+            } => {
+                assert_eq!(commits_behind, 2);
+                assert_eq!(missing_fixes.len(), 2);
+                assert_eq!(missing_fixes[0], "fix: handle null pointer");
+                assert_eq!(missing_fixes[1], "fix: resolve timeout");
+            }
+            other => panic!("expected Stale, got {other:?}"),
+        }
+
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn diverged_branch_detection() {
+        let root = temp_dir();
+        init_repo(&root);
+
+        // given
+        run(&root, &["checkout", "-b", "topic"]);
+        commit_file(&root, "topic_work.txt", "topic work");
+        run(&root, &["checkout", "main"]);
+        commit_file(&root, "main_fix.txt", "main fix");
+
+        // when
+        let freshness = check_freshness_in("topic", "main", &root);
+
+        // then
+        match freshness {
+            BranchFreshness::Diverged {
+                ahead,
+                behind,
+                missing_fixes,
+            } => {
+                assert_eq!(ahead, 1);
+                assert_eq!(behind, 1);
+                assert_eq!(missing_fixes, vec!["main fix".to_string()]);
+            }
+            other => panic!("expected Diverged, got {other:?}"),
+        }
+
+        fs::remove_dir_all(&root).expect("cleanup");
+    }
+
+    #[test]
+    fn policy_noop_for_fresh_branch() {
+        // given
+        let freshness = BranchFreshness::Fresh;
+
+        // when
+        let action = apply_policy(&freshness, StaleBranchPolicy::WarnOnly);
+
+        // then
+        assert_eq!(action, StaleBranchAction::Noop);
+    }
+
+    #[test]
+    fn policy_warn_for_stale_branch() {
+        // given
+        let freshness = BranchFreshness::Stale {
+            commits_behind: 3,
+            missing_fixes: vec!["fix: timeout".into(), "fix: null ptr".into()],
+        };
+
+        // when
+        let action = apply_policy(&freshness, StaleBranchPolicy::WarnOnly);
+
+        // then
+        match action {
+            StaleBranchAction::Warn { message } => {
+                assert!(message.contains("3 commit(s) behind"));
+                assert!(message.contains("fix: timeout"));
+                assert!(message.contains("fix: null ptr"));
+            }
+            other => panic!("expected Warn, got {other:?}"),
+        }
+    }
+
+    #[test]
+    fn policy_block_for_stale_branch() {
+        // given
+        let freshness = BranchFreshness::Stale {
+            commits_behind: 1,
+            missing_fixes: vec!["hotfix".into()],
+        };
+
+        // when
+        let action = apply_policy(&freshness, StaleBranchPolicy::Block);
+
+        // then
+        match action {
+            StaleBranchAction::Block { message } => {
+                assert!(message.contains("1 commit(s) behind"));
+            }
+            other => panic!("expected Block, got {other:?}"),
+        }
+    }
+
+    #[test]
+    fn policy_auto_rebase_for_stale_branch() {
+        // given
+        let freshness = BranchFreshness::Stale {
+            commits_behind: 2,
+            missing_fixes: vec![],
+        };
+
+        // when
+        let action = apply_policy(&freshness, StaleBranchPolicy::AutoRebase);
+
+        // then
+        assert_eq!(action, StaleBranchAction::Rebase);
+    }
+
+    #[test]
+    fn policy_auto_merge_forward_for_diverged_branch() {
+        // given
+        let freshness = BranchFreshness::Diverged {
+            ahead: 5,
+            behind: 2,
+            missing_fixes: vec!["fix: merge main".into()],
+        };
+
+        // when
+        let action = apply_policy(&freshness, StaleBranchPolicy::AutoMergeForward);
+
+        // then
+        assert_eq!(action, StaleBranchAction::MergeForward);
+    }
+
+    #[test]
+    fn policy_warn_for_diverged_branch() {
+        // given
+        let freshness = BranchFreshness::Diverged {
+            ahead: 3,
+            behind: 1,
+            missing_fixes: vec!["main hotfix".into()],
+        };
+
+        // when
+        let action = apply_policy(&freshness, StaleBranchPolicy::WarnOnly);
+
+        // then
+        match action {
+            StaleBranchAction::Warn { message } => {
+                assert!(message.contains("diverged"));
+                assert!(message.contains("3 commit(s) ahead"));
+                assert!(message.contains("1 commit(s) behind"));
+                assert!(message.contains("main hotfix"));
+            }
+            other => panic!("expected Warn, got {other:?}"),
+        }
+    }
+}
@@ -0,0 +1,300 @@
+use std::collections::BTreeSet;
+
+const DEFAULT_MAX_CHARS: usize = 1_200;
+const DEFAULT_MAX_LINES: usize = 24;
+const DEFAULT_MAX_LINE_CHARS: usize = 160;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub struct SummaryCompressionBudget {
+    pub max_chars: usize,
+    pub max_lines: usize,
+    pub max_line_chars: usize,
+}
+
+impl Default for SummaryCompressionBudget {
+    fn default() -> Self {
+        Self {
+            max_chars: DEFAULT_MAX_CHARS,
+            max_lines: DEFAULT_MAX_LINES,
+            max_line_chars: DEFAULT_MAX_LINE_CHARS,
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct SummaryCompressionResult {
+    pub summary: String,
+    pub original_chars: usize,
+    pub compressed_chars: usize,
+    pub original_lines: usize,
+    pub compressed_lines: usize,
+    pub removed_duplicate_lines: usize,
+    pub omitted_lines: usize,
+    pub truncated: bool,
+}
+
+#[must_use]
+pub fn compress_summary(
+    summary: &str,
+    budget: SummaryCompressionBudget,
+) -> SummaryCompressionResult {
+    let original_chars = summary.chars().count();
+    let original_lines = summary.lines().count();
+
+    let normalized = normalize_lines(summary, budget.max_line_chars);
+    if normalized.lines.is_empty() || budget.max_chars == 0 || budget.max_lines == 0 {
+        return SummaryCompressionResult {
+            summary: String::new(),
+            original_chars,
+            compressed_chars: 0,
+            original_lines,
+            compressed_lines: 0,
+            removed_duplicate_lines: normalized.removed_duplicate_lines,
+            omitted_lines: normalized.lines.len(),
+            truncated: original_chars > 0,
+        };
+    }
+
+    let selected = select_line_indexes(&normalized.lines, budget);
+    let mut compressed_lines = selected
+        .iter()
+        .map(|index| normalized.lines[*index].clone())
+        .collect::<Vec<_>>();
+    if compressed_lines.is_empty() {
+        compressed_lines.push(truncate_line(&normalized.lines[0], budget.max_chars));
+    }
+    let omitted_lines = normalized
+        .lines
+        .len()
+        .saturating_sub(compressed_lines.len());
+
+    if omitted_lines > 0 {
+        let omission_notice = omission_notice(omitted_lines);
+        push_line_with_budget(&mut compressed_lines, omission_notice, budget);
+    }
+
+    let compressed_summary = compressed_lines.join("\n");
+
+    SummaryCompressionResult {
+        compressed_chars: compressed_summary.chars().count(),
+        compressed_lines: compressed_lines.len(),
+        removed_duplicate_lines: normalized.removed_duplicate_lines,
+        omitted_lines,
+        truncated: compressed_summary != summary.trim(),
+        summary: compressed_summary,
+        original_chars,
+        original_lines,
+    }
+}
+
+#[must_use]
+pub fn compress_summary_text(summary: &str) -> String {
+    compress_summary(summary, SummaryCompressionBudget::default()).summary
+}
+
+#[derive(Debug, Default)]
+struct NormalizedSummary {
+    lines: Vec<String>,
+    removed_duplicate_lines: usize,
+}
+
+fn normalize_lines(summary: &str, max_line_chars: usize) -> NormalizedSummary {
+    let mut seen = BTreeSet::new();
+    let mut lines = Vec::new();
+    let mut removed_duplicate_lines = 0;
+
+    for raw_line in summary.lines() {
+        let normalized = collapse_inline_whitespace(raw_line);
+        if normalized.is_empty() {
+            continue;
+        }
+
+        let truncated = truncate_line(&normalized, max_line_chars);
+        let dedupe_key = dedupe_key(&truncated);
+        if !seen.insert(dedupe_key) {
+            removed_duplicate_lines += 1;
+            continue;
+        }
+
+        lines.push(truncated);
+    }
+
+    NormalizedSummary {
+        lines,
+        removed_duplicate_lines,
+    }
+}
+
+fn select_line_indexes(lines: &[String], budget: SummaryCompressionBudget) -> Vec<usize> {
+    let mut selected = BTreeSet::<usize>::new();
+
+    for priority in 0..=3 {
+        for (index, line) in lines.iter().enumerate() {
+            if selected.contains(&index) || line_priority(line) != priority {
+                continue;
+            }
+
+            let candidate = selected
+                .iter()
+                .map(|selected_index| lines[*selected_index].as_str())
+                .chain(std::iter::once(line.as_str()))
+                .collect::<Vec<_>>();
+
+            if candidate.len() > budget.max_lines {
+                continue;
+            }
+
+            if joined_char_count(&candidate) > budget.max_chars {
+                continue;
+            }
+
+            selected.insert(index);
+        }
+    }
+
+    selected.into_iter().collect()
+}
+
+fn push_line_with_budget(lines: &mut Vec<String>, line: String, budget: SummaryCompressionBudget) {
+    let candidate = lines
+        .iter()
+        .map(String::as_str)
+        .chain(std::iter::once(line.as_str()))
+        .collect::<Vec<_>>();
+
+    if candidate.len() <= budget.max_lines && joined_char_count(&candidate) <= budget.max_chars {
+        lines.push(line);
+    }
+}
+
+fn joined_char_count(lines: &[&str]) -> usize {
+    lines.iter().map(|line| line.chars().count()).sum::<usize>() + lines.len().saturating_sub(1)
+}
+
+fn line_priority(line: &str) -> usize {
+    if line == "Summary:" || line == "Conversation summary:" || is_core_detail(line) {
+        0
+    } else if is_section_header(line) {
+        1
+    } else if line.starts_with("- ") || line.starts_with("  - ") {
+        2
+    } else {
+        3
+    }
+}
+
+fn is_core_detail(line: &str) -> bool {
+    [
+        "- Scope:",
+        "- Current work:",
+        "- Pending work:",
+        "- Key files referenced:",
+        "- Tools mentioned:",
+        "- Recent user requests:",
+        "- Previously compacted context:",
+        "- Newly compacted context:",
+    ]
+    .iter()
+    .any(|prefix| line.starts_with(prefix))
+}
+
+fn is_section_header(line: &str) -> bool {
+    line.ends_with(':')
+}
+
+fn omission_notice(omitted_lines: usize) -> String {
+    format!("- … {omitted_lines} additional line(s) omitted.")
+}
+
+fn collapse_inline_whitespace(line: &str) -> String {
+    line.split_whitespace().collect::<Vec<_>>().join(" ")
+}
+
+fn truncate_line(line: &str, max_chars: usize) -> String {
+    if max_chars == 0 || line.chars().count() <= max_chars {
+        return line.to_string();
+    }
+
+    if max_chars == 1 {
+        return "…".to_string();
+    }
+
+    let mut truncated = line
+        .chars()
+        .take(max_chars.saturating_sub(1))
+        .collect::<String>();
+    truncated.push('…');
+    truncated
+}
+
+fn dedupe_key(line: &str) -> String {
+    line.to_ascii_lowercase()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{compress_summary, compress_summary_text, SummaryCompressionBudget};
+
+    #[test]
+    fn collapses_whitespace_and_duplicate_lines() {
+        // given
+        let summary = "Conversation summary:\n\n- Scope:   compact   earlier   messages.\n- Scope: compact earlier messages.\n- Current work: update runtime module.\n";
+
+        // when
+        let result = compress_summary(summary, SummaryCompressionBudget::default());
+
+        // then
+        assert_eq!(result.removed_duplicate_lines, 1);
+        assert!(result
+            .summary
+            .contains("- Scope: compact earlier messages."));
+        assert!(!result.summary.contains("  compact   earlier"));
+    }
+
+    #[test]
+    fn keeps_core_lines_when_budget_is_tight() {
+        // given
+        let summary = [
+            "Conversation summary:",
+            "- Scope: 18 earlier messages compacted.",
+            "- Current work: finish summary compression.",
+            "- Key timeline:",
+            "  - user: asked for a working implementation.",
+            "  - assistant: inspected runtime compaction flow.",
+            "  - tool: cargo check succeeded.",
+        ]
+        .join("\n");
+
+        // when
+        let result = compress_summary(
+            &summary,
+            SummaryCompressionBudget {
+                max_chars: 120,
+                max_lines: 3,
+                max_line_chars: 80,
+            },
+        );
+
+        // then
+        assert!(result.summary.contains("Conversation summary:"));
+        assert!(result
+            .summary
+            .contains("- Scope: 18 earlier messages compacted."));
+        assert!(result
+            .summary
+            .contains("- Current work: finish summary compression."));
+        assert!(result.omitted_lines > 0);
+    }
+
+    #[test]
+    fn provides_a_default_text_only_helper() {
+        // given
+        let summary = "Summary:\n\nA short line.";
+
+        // when
+        let compressed = compress_summary_text(summary);
+
+        // then
+        assert_eq!(compressed, "Summary:\nA short line.");
+    }
+}
@@ -0,0 +1,158 @@
+use serde::{Deserialize, Serialize};
+use std::fmt::{Display, Formatter};
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+pub struct TaskPacket {
+    pub objective: String,
+    pub scope: String,
+    pub repo: String,
+    pub branch_policy: String,
+    pub acceptance_tests: Vec<String>,
+    pub commit_policy: String,
+    pub reporting_contract: String,
+    pub escalation_policy: String,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct TaskPacketValidationError {
+    errors: Vec<String>,
+}
+
+impl TaskPacketValidationError {
+    #[must_use]
+    pub fn new(errors: Vec<String>) -> Self {
+        Self { errors }
+    }
+
+    #[must_use]
+    pub fn errors(&self) -> &[String] {
+        &self.errors
+    }
+}
+
+impl Display for TaskPacketValidationError {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", self.errors.join("; "))
+    }
+}
+
+impl std::error::Error for TaskPacketValidationError {}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ValidatedPacket(TaskPacket);
+
+impl ValidatedPacket {
+    #[must_use]
+    pub fn packet(&self) -> &TaskPacket {
+        &self.0
+    }
+
+    #[must_use]
+    pub fn into_inner(self) -> TaskPacket {
+        self.0
+    }
+}
+
+pub fn validate_packet(packet: TaskPacket) -> Result<ValidatedPacket, TaskPacketValidationError> {
+    let mut errors = Vec::new();
+
+    validate_required("objective", &packet.objective, &mut errors);
+    validate_required("scope", &packet.scope, &mut errors);
+    validate_required("repo", &packet.repo, &mut errors);
+    validate_required("branch_policy", &packet.branch_policy, &mut errors);
+    validate_required("commit_policy", &packet.commit_policy, &mut errors);
+    validate_required(
+        "reporting_contract",
+        &packet.reporting_contract,
+        &mut errors,
+    );
+    validate_required("escalation_policy", &packet.escalation_policy, &mut errors);
+
+    for (index, test) in packet.acceptance_tests.iter().enumerate() {
+        if test.trim().is_empty() {
+            errors.push(format!(
+                "acceptance_tests contains an empty value at index {index}"
+            ));
+        }
+    }
+
+    if errors.is_empty() {
+        Ok(ValidatedPacket(packet))
+    } else {
+        Err(TaskPacketValidationError::new(errors))
+    }
+}
+
+fn validate_required(field: &str, value: &str, errors: &mut Vec<String>) {
+    if value.trim().is_empty() {
+        errors.push(format!("{field} must not be empty"));
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    fn sample_packet() -> TaskPacket {
+        TaskPacket {
+            objective: "Implement typed task packet format".to_string(),
+            scope: "runtime/task system".to_string(),
+            repo: "claw-code-parity".to_string(),
+            branch_policy: "origin/main only".to_string(),
+            acceptance_tests: vec![
+                "cargo build --workspace".to_string(),
+                "cargo test --workspace".to_string(),
+            ],
+            commit_policy: "single verified commit".to_string(),
+            reporting_contract: "print build result, test result, commit sha".to_string(),
+            escalation_policy: "stop only on destructive ambiguity".to_string(),
+        }
+    }
+
+    #[test]
+    fn valid_packet_passes_validation() {
+        let packet = sample_packet();
+        let validated = validate_packet(packet.clone()).expect("packet should validate");
+        assert_eq!(validated.packet(), &packet);
+        assert_eq!(validated.into_inner(), packet);
+    }
+
+    #[test]
+    fn invalid_packet_accumulates_errors() {
+        let packet = TaskPacket {
+            objective: " ".to_string(),
+            scope: String::new(),
+            repo: String::new(),
+            branch_policy: "\t".to_string(),
+            acceptance_tests: vec!["ok".to_string(), " ".to_string()],
+            commit_policy: String::new(),
+            reporting_contract: String::new(),
+            escalation_policy: String::new(),
+        };
+
+        let error = validate_packet(packet).expect_err("packet should be rejected");
+
+        assert!(error.errors().len() >= 7);
+        assert!(error
+            .errors()
+            .contains(&"objective must not be empty".to_string()));
+        assert!(error
+            .errors()
+            .contains(&"scope must not be empty".to_string()));
+        assert!(error
+            .errors()
+            .contains(&"repo must not be empty".to_string()));
+        assert!(error
+            .errors()
+            .contains(&"acceptance_tests contains an empty value at index 1".to_string()));
+    }
+
+    #[test]
+    fn serialization_roundtrip_preserves_packet() {
+        let packet = sample_packet();
+        let serialized = serde_json::to_string(&packet).expect("packet should serialize");
+        let deserialized: TaskPacket =
+            serde_json::from_str(&serialized).expect("packet should deserialize");
+        assert_eq!(deserialized, packet);
+    }
+}
@@ -0,0 +1,503 @@
+#![allow(clippy::must_use_candidate, clippy::unnecessary_map_or)]
+//! In-memory task registry for sub-agent task lifecycle management.
+
+use std::collections::HashMap;
+use std::sync::{Arc, Mutex};
+use std::time::{SystemTime, UNIX_EPOCH};
+
+use serde::{Deserialize, Serialize};
+
+use crate::{validate_packet, TaskPacket, TaskPacketValidationError};
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum TaskStatus {
+    Created,
+    Running,
+    Completed,
+    Failed,
+    Stopped,
+}
+
+impl std::fmt::Display for TaskStatus {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Created => write!(f, "created"),
+            Self::Running => write!(f, "running"),
+            Self::Completed => write!(f, "completed"),
+            Self::Failed => write!(f, "failed"),
+            Self::Stopped => write!(f, "stopped"),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct Task {
+    pub task_id: String,
+    pub prompt: String,
+    pub description: Option<String>,
+    pub task_packet: Option<TaskPacket>,
+    pub status: TaskStatus,
+    pub created_at: u64,
+    pub updated_at: u64,
+    pub messages: Vec<TaskMessage>,
+    pub output: String,
+    pub team_id: Option<String>,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct TaskMessage {
+    pub role: String,
+    pub content: String,
+    pub timestamp: u64,
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct TaskRegistry {
+    inner: Arc<Mutex<RegistryInner>>,
+}
+
+#[derive(Debug, Default)]
+struct RegistryInner {
+    tasks: HashMap<String, Task>,
+    counter: u64,
+}
+
+fn now_secs() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .unwrap_or_default()
+        .as_secs()
+}
+
+impl TaskRegistry {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    pub fn create(&self, prompt: &str, description: Option<&str>) -> Task {
+        self.create_task(prompt.to_owned(), description.map(str::to_owned), None)
+    }
+
+    pub fn create_from_packet(
+        &self,
+        packet: TaskPacket,
+    ) -> Result<Task, TaskPacketValidationError> {
+        let packet = validate_packet(packet)?.into_inner();
+        Ok(self.create_task(
+            packet.objective.clone(),
+            Some(packet.scope.clone()),
+            Some(packet),
+        ))
+    }
+
+    fn create_task(
+        &self,
+        prompt: String,
+        description: Option<String>,
+        task_packet: Option<TaskPacket>,
+    ) -> Task {
+        let mut inner = self.inner.lock().expect("registry lock poisoned");
+        inner.counter += 1;
+        let ts = now_secs();
+        let task_id = format!("task_{:08x}_{}", ts, inner.counter);
+        let task = Task {
+            task_id: task_id.clone(),
+            prompt,
+            description,
+            task_packet,
+            status: TaskStatus::Created,
+            created_at: ts,
+            updated_at: ts,
+            messages: Vec::new(),
+            output: String::new(),
+            team_id: None,
+        };
+        inner.tasks.insert(task_id, task.clone());
+        task
+    }
+
+    pub fn get(&self, task_id: &str) -> Option<Task> {
+        let inner = self.inner.lock().expect("registry lock poisoned");
+        inner.tasks.get(task_id).cloned()
+    }
+
+    pub fn list(&self, status_filter: Option<TaskStatus>) -> Vec<Task> {
+        let inner = self.inner.lock().expect("registry lock poisoned");
+        inner
+            .tasks
+            .values()
+            .filter(|t| status_filter.map_or(true, |s| t.status == s))
+            .cloned()
+            .collect()
+    }
+
+    pub fn stop(&self, task_id: &str) -> Result<Task, String> {
+        let mut inner = self.inner.lock().expect("registry lock poisoned");
+        let task = inner
+            .tasks
+            .get_mut(task_id)
+            .ok_or_else(|| format!("task not found: {task_id}"))?;
+
+        match task.status {
+            TaskStatus::Completed | TaskStatus::Failed | TaskStatus::Stopped => {
+                return Err(format!(
+                    "task {task_id} is already in terminal state: {}",
+                    task.status
+                ));
+            }
+            _ => {}
+        }
+
+        task.status = TaskStatus::Stopped;
+        task.updated_at = now_secs();
+        Ok(task.clone())
+    }
+
+    pub fn update(&self, task_id: &str, message: &str) -> Result<Task, String> {
+        let mut inner = self.inner.lock().expect("registry lock poisoned");
+        let task = inner
+            .tasks
+            .get_mut(task_id)
+            .ok_or_else(|| format!("task not found: {task_id}"))?;
+
+        task.messages.push(TaskMessage {
+            role: String::from("user"),
+            content: message.to_owned(),
+            timestamp: now_secs(),
+        });
+        task.updated_at = now_secs();
+        Ok(task.clone())
+    }
+
+    pub fn output(&self, task_id: &str) -> Result<String, String> {
+        let inner = self.inner.lock().expect("registry lock poisoned");
+        let task = inner
+            .tasks
+            .get(task_id)
+            .ok_or_else(|| format!("task not found: {task_id}"))?;
+        Ok(task.output.clone())
+    }
+
+    pub fn append_output(&self, task_id: &str, output: &str) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("registry lock poisoned");
+        let task = inner
+            .tasks
+            .get_mut(task_id)
+            .ok_or_else(|| format!("task not found: {task_id}"))?;
+        task.output.push_str(output);
+        task.updated_at = now_secs();
+        Ok(())
+    }
+
+    pub fn set_status(&self, task_id: &str, status: TaskStatus) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("registry lock poisoned");
+        let task = inner
+            .tasks
+            .get_mut(task_id)
+            .ok_or_else(|| format!("task not found: {task_id}"))?;
+        task.status = status;
+        task.updated_at = now_secs();
+        Ok(())
+    }
+
+    pub fn assign_team(&self, task_id: &str, team_id: &str) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("registry lock poisoned");
+        let task = inner
+            .tasks
+            .get_mut(task_id)
+            .ok_or_else(|| format!("task not found: {task_id}"))?;
+        task.team_id = Some(team_id.to_owned());
+        task.updated_at = now_secs();
+        Ok(())
+    }
+
+    pub fn remove(&self, task_id: &str) -> Option<Task> {
+        let mut inner = self.inner.lock().expect("registry lock poisoned");
+        inner.tasks.remove(task_id)
+    }
+
+    #[must_use]
+    pub fn len(&self) -> usize {
+        let inner = self.inner.lock().expect("registry lock poisoned");
+        inner.tasks.len()
+    }
+
+    #[must_use]
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn creates_and_retrieves_tasks() {
+        let registry = TaskRegistry::new();
+        let task = registry.create("Do something", Some("A test task"));
+        assert_eq!(task.status, TaskStatus::Created);
+        assert_eq!(task.prompt, "Do something");
+        assert_eq!(task.description.as_deref(), Some("A test task"));
+        assert_eq!(task.task_packet, None);
+
+        let fetched = registry.get(&task.task_id).expect("task should exist");
+        assert_eq!(fetched.task_id, task.task_id);
+    }
+
+    #[test]
+    fn creates_task_from_packet() {
+        let registry = TaskRegistry::new();
+        let packet = TaskPacket {
+            objective: "Ship task packet support".to_string(),
+            scope: "runtime/task system".to_string(),
+            repo: "claw-code-parity".to_string(),
+            branch_policy: "origin/main only".to_string(),
+            acceptance_tests: vec!["cargo test --workspace".to_string()],
+            commit_policy: "single commit".to_string(),
+            reporting_contract: "print commit sha".to_string(),
+            escalation_policy: "manual escalation".to_string(),
+        };
+
+        let task = registry
+            .create_from_packet(packet.clone())
+            .expect("packet-backed task should be created");
+
+        assert_eq!(task.prompt, packet.objective);
+        assert_eq!(task.description.as_deref(), Some("runtime/task system"));
+        assert_eq!(task.task_packet, Some(packet.clone()));
+
+        let fetched = registry.get(&task.task_id).expect("task should exist");
+        assert_eq!(fetched.task_packet, Some(packet));
+    }
+
+    #[test]
+    fn lists_tasks_with_optional_filter() {
+        let registry = TaskRegistry::new();
+        registry.create("Task A", None);
+        let task_b = registry.create("Task B", None);
+        registry
+            .set_status(&task_b.task_id, TaskStatus::Running)
+            .expect("set status should succeed");
+
+        let all = registry.list(None);
+        assert_eq!(all.len(), 2);
+
+        let running = registry.list(Some(TaskStatus::Running));
+        assert_eq!(running.len(), 1);
+        assert_eq!(running[0].task_id, task_b.task_id);
+
+        let created = registry.list(Some(TaskStatus::Created));
+        assert_eq!(created.len(), 1);
+    }
+
+    #[test]
+    fn stops_running_task() {
+        let registry = TaskRegistry::new();
+        let task = registry.create("Stoppable", None);
+        registry
+            .set_status(&task.task_id, TaskStatus::Running)
+            .unwrap();
+
+        let stopped = registry.stop(&task.task_id).expect("stop should succeed");
+        assert_eq!(stopped.status, TaskStatus::Stopped);
+
+        // Stopping again should fail
+        let result = registry.stop(&task.task_id);
+        assert!(result.is_err());
+    }
+
+    #[test]
+    fn updates_task_with_messages() {
+        let registry = TaskRegistry::new();
+        let task = registry.create("Messageable", None);
+        let updated = registry
+            .update(&task.task_id, "Here's more context")
+            .expect("update should succeed");
+        assert_eq!(updated.messages.len(), 1);
+        assert_eq!(updated.messages[0].content, "Here's more context");
+        assert_eq!(updated.messages[0].role, "user");
+    }
+
+    #[test]
+    fn appends_and_retrieves_output() {
+        let registry = TaskRegistry::new();
+        let task = registry.create("Output task", None);
+        registry
+            .append_output(&task.task_id, "line 1\n")
+            .expect("append should succeed");
+        registry
+            .append_output(&task.task_id, "line 2\n")
+            .expect("append should succeed");
+
+        let output = registry.output(&task.task_id).expect("output should exist");
+        assert_eq!(output, "line 1\nline 2\n");
+    }
+
+    #[test]
+    fn assigns_team_and_removes_task() {
+        let registry = TaskRegistry::new();
+        let task = registry.create("Team task", None);
+        registry
+            .assign_team(&task.task_id, "team_abc")
+            .expect("assign should succeed");
+
+        let fetched = registry.get(&task.task_id).unwrap();
+        assert_eq!(fetched.team_id.as_deref(), Some("team_abc"));
+
+        let removed = registry.remove(&task.task_id);
+        assert!(removed.is_some());
+        assert!(registry.get(&task.task_id).is_none());
+        assert!(registry.is_empty());
+    }
+
+    #[test]
+    fn rejects_operations_on_missing_task() {
+        let registry = TaskRegistry::new();
+        assert!(registry.stop("nonexistent").is_err());
+        assert!(registry.update("nonexistent", "msg").is_err());
+        assert!(registry.output("nonexistent").is_err());
+        assert!(registry.append_output("nonexistent", "data").is_err());
+        assert!(registry
+            .set_status("nonexistent", TaskStatus::Running)
+            .is_err());
+    }
+
+    #[test]
+    fn task_status_display_all_variants() {
+        // given
+        let cases = [
+            (TaskStatus::Created, "created"),
+            (TaskStatus::Running, "running"),
+            (TaskStatus::Completed, "completed"),
+            (TaskStatus::Failed, "failed"),
+            (TaskStatus::Stopped, "stopped"),
+        ];
+
+        // when
+        let rendered: Vec<_> = cases
+            .into_iter()
+            .map(|(status, expected)| (status.to_string(), expected))
+            .collect();
+
+        // then
+        assert_eq!(
+            rendered,
+            vec![
+                ("created".to_string(), "created"),
+                ("running".to_string(), "running"),
+                ("completed".to_string(), "completed"),
+                ("failed".to_string(), "failed"),
+                ("stopped".to_string(), "stopped"),
+            ]
+        );
+    }
+
+    #[test]
+    fn stop_rejects_completed_task() {
+        // given
+        let registry = TaskRegistry::new();
+        let task = registry.create("done", None);
+        registry
+            .set_status(&task.task_id, TaskStatus::Completed)
+            .expect("set status should succeed");
+
+        // when
+        let result = registry.stop(&task.task_id);
+
+        // then
+        let error = result.expect_err("completed task should be rejected");
+        assert!(error.contains("already in terminal state"));
+        assert!(error.contains("completed"));
+    }
+
+    #[test]
+    fn stop_rejects_failed_task() {
+        // given
+        let registry = TaskRegistry::new();
+        let task = registry.create("failed", None);
+        registry
+            .set_status(&task.task_id, TaskStatus::Failed)
+            .expect("set status should succeed");
+
+        // when
+        let result = registry.stop(&task.task_id);
+
+        // then
+        let error = result.expect_err("failed task should be rejected");
+        assert!(error.contains("already in terminal state"));
+        assert!(error.contains("failed"));
+    }
+
+    #[test]
+    fn stop_succeeds_from_created_state() {
+        // given
+        let registry = TaskRegistry::new();
+        let task = registry.create("created task", None);
+
+        // when
+        let stopped = registry.stop(&task.task_id).expect("stop should succeed");
+
+        // then
+        assert_eq!(stopped.status, TaskStatus::Stopped);
+        assert!(stopped.updated_at >= task.updated_at);
+    }
+
+    #[test]
+    fn new_registry_is_empty() {
+        // given
+        let registry = TaskRegistry::new();
+
+        // when
+        let all_tasks = registry.list(None);
+
+        // then
+        assert!(registry.is_empty());
+        assert_eq!(registry.len(), 0);
+        assert!(all_tasks.is_empty());
+    }
+
+    #[test]
+    fn create_without_description() {
+        // given
+        let registry = TaskRegistry::new();
+
+        // when
+        let task = registry.create("Do the thing", None);
+
+        // then
+        assert!(task.task_id.starts_with("task_"));
+        assert_eq!(task.description, None);
+        assert_eq!(task.task_packet, None);
+        assert!(task.messages.is_empty());
+        assert!(task.output.is_empty());
+        assert_eq!(task.team_id, None);
+    }
+
+    #[test]
+    fn remove_nonexistent_returns_none() {
+        // given
+        let registry = TaskRegistry::new();
+
+        // when
+        let removed = registry.remove("missing");
+
+        // then
+        assert!(removed.is_none());
+    }
+
+    #[test]
+    fn assign_team_rejects_missing_task() {
+        // given
+        let registry = TaskRegistry::new();
+
+        // when
+        let result = registry.assign_team("missing", "team_123");
+
+        // then
+        let error = result.expect_err("missing task should be rejected");
+        assert_eq!(error, "task not found: missing");
+    }
+}
@@ -0,0 +1,509 @@
+#![allow(clippy::must_use_candidate)]
+//! In-memory registries for Team and Cron lifecycle management.
+//!
+//! Provides TeamCreate/Delete and CronCreate/Delete/List runtime backing
+//! to replace the stub implementations in the tools crate.
+
+use std::collections::HashMap;
+use std::sync::{Arc, Mutex};
+use std::time::{SystemTime, UNIX_EPOCH};
+
+use serde::{Deserialize, Serialize};
+
+fn now_secs() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .unwrap_or_default()
+        .as_secs()
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct Team {
+    pub team_id: String,
+    pub name: String,
+    pub task_ids: Vec<String>,
+    pub status: TeamStatus,
+    pub created_at: u64,
+    pub updated_at: u64,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum TeamStatus {
+    Created,
+    Running,
+    Completed,
+    Deleted,
+}
+
+impl std::fmt::Display for TeamStatus {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::Created => write!(f, "created"),
+            Self::Running => write!(f, "running"),
+            Self::Completed => write!(f, "completed"),
+            Self::Deleted => write!(f, "deleted"),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct TeamRegistry {
+    inner: Arc<Mutex<TeamInner>>,
+}
+
+#[derive(Debug, Default)]
+struct TeamInner {
+    teams: HashMap<String, Team>,
+    counter: u64,
+}
+
+impl TeamRegistry {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    pub fn create(&self, name: &str, task_ids: Vec<String>) -> Team {
+        let mut inner = self.inner.lock().expect("team registry lock poisoned");
+        inner.counter += 1;
+        let ts = now_secs();
+        let team_id = format!("team_{:08x}_{}", ts, inner.counter);
+        let team = Team {
+            team_id: team_id.clone(),
+            name: name.to_owned(),
+            task_ids,
+            status: TeamStatus::Created,
+            created_at: ts,
+            updated_at: ts,
+        };
+        inner.teams.insert(team_id, team.clone());
+        team
+    }
+
+    pub fn get(&self, team_id: &str) -> Option<Team> {
+        let inner = self.inner.lock().expect("team registry lock poisoned");
+        inner.teams.get(team_id).cloned()
+    }
+
+    pub fn list(&self) -> Vec<Team> {
+        let inner = self.inner.lock().expect("team registry lock poisoned");
+        inner.teams.values().cloned().collect()
+    }
+
+    pub fn delete(&self, team_id: &str) -> Result<Team, String> {
+        let mut inner = self.inner.lock().expect("team registry lock poisoned");
+        let team = inner
+            .teams
+            .get_mut(team_id)
+            .ok_or_else(|| format!("team not found: {team_id}"))?;
+        team.status = TeamStatus::Deleted;
+        team.updated_at = now_secs();
+        Ok(team.clone())
+    }
+
+    pub fn remove(&self, team_id: &str) -> Option<Team> {
+        let mut inner = self.inner.lock().expect("team registry lock poisoned");
+        inner.teams.remove(team_id)
+    }
+
+    #[must_use]
+    pub fn len(&self) -> usize {
+        let inner = self.inner.lock().expect("team registry lock poisoned");
+        inner.teams.len()
+    }
+
+    #[must_use]
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct CronEntry {
+    pub cron_id: String,
+    pub schedule: String,
+    pub prompt: String,
+    pub description: Option<String>,
+    pub enabled: bool,
+    pub created_at: u64,
+    pub updated_at: u64,
+    pub last_run_at: Option<u64>,
+    pub run_count: u64,
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct CronRegistry {
+    inner: Arc<Mutex<CronInner>>,
+}
+
+#[derive(Debug, Default)]
+struct CronInner {
+    entries: HashMap<String, CronEntry>,
+    counter: u64,
+}
+
+impl CronRegistry {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    pub fn create(&self, schedule: &str, prompt: &str, description: Option<&str>) -> CronEntry {
+        let mut inner = self.inner.lock().expect("cron registry lock poisoned");
+        inner.counter += 1;
+        let ts = now_secs();
+        let cron_id = format!("cron_{:08x}_{}", ts, inner.counter);
+        let entry = CronEntry {
+            cron_id: cron_id.clone(),
+            schedule: schedule.to_owned(),
+            prompt: prompt.to_owned(),
+            description: description.map(str::to_owned),
+            enabled: true,
+            created_at: ts,
+            updated_at: ts,
+            last_run_at: None,
+            run_count: 0,
+        };
+        inner.entries.insert(cron_id, entry.clone());
+        entry
+    }
+
+    pub fn get(&self, cron_id: &str) -> Option<CronEntry> {
+        let inner = self.inner.lock().expect("cron registry lock poisoned");
+        inner.entries.get(cron_id).cloned()
+    }
+
+    pub fn list(&self, enabled_only: bool) -> Vec<CronEntry> {
+        let inner = self.inner.lock().expect("cron registry lock poisoned");
+        inner
+            .entries
+            .values()
+            .filter(|e| !enabled_only || e.enabled)
+            .cloned()
+            .collect()
+    }
+
+    pub fn delete(&self, cron_id: &str) -> Result<CronEntry, String> {
+        let mut inner = self.inner.lock().expect("cron registry lock poisoned");
+        inner
+            .entries
+            .remove(cron_id)
+            .ok_or_else(|| format!("cron not found: {cron_id}"))
+    }
+
+    /// Disable a cron entry without removing it.
+    pub fn disable(&self, cron_id: &str) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("cron registry lock poisoned");
+        let entry = inner
+            .entries
+            .get_mut(cron_id)
+            .ok_or_else(|| format!("cron not found: {cron_id}"))?;
+        entry.enabled = false;
+        entry.updated_at = now_secs();
+        Ok(())
+    }
+
+    /// Record a cron run.
+    pub fn record_run(&self, cron_id: &str) -> Result<(), String> {
+        let mut inner = self.inner.lock().expect("cron registry lock poisoned");
+        let entry = inner
+            .entries
+            .get_mut(cron_id)
+            .ok_or_else(|| format!("cron not found: {cron_id}"))?;
+        entry.last_run_at = Some(now_secs());
+        entry.run_count += 1;
+        entry.updated_at = now_secs();
+        Ok(())
+    }
+
+    #[must_use]
+    pub fn len(&self) -> usize {
+        let inner = self.inner.lock().expect("cron registry lock poisoned");
+        inner.entries.len()
+    }
+
+    #[must_use]
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    // ── Team tests ──────────────────────────────────────
+
+    #[test]
+    fn creates_and_retrieves_team() {
+        let registry = TeamRegistry::new();
+        let team = registry.create("Alpha Squad", vec!["task_001".into(), "task_002".into()]);
+        assert_eq!(team.name, "Alpha Squad");
+        assert_eq!(team.task_ids.len(), 2);
+        assert_eq!(team.status, TeamStatus::Created);
+
+        let fetched = registry.get(&team.team_id).expect("team should exist");
+        assert_eq!(fetched.team_id, team.team_id);
+    }
+
+    #[test]
+    fn lists_and_deletes_teams() {
+        let registry = TeamRegistry::new();
+        let t1 = registry.create("Team A", vec![]);
+        let t2 = registry.create("Team B", vec![]);
+
+        let all = registry.list();
+        assert_eq!(all.len(), 2);
+
+        let deleted = registry.delete(&t1.team_id).expect("delete should succeed");
+        assert_eq!(deleted.status, TeamStatus::Deleted);
+
+        // Team is still listable (soft delete)
+        let still_there = registry.get(&t1.team_id).unwrap();
+        assert_eq!(still_there.status, TeamStatus::Deleted);
+
+        // Hard remove
+        registry.remove(&t2.team_id);
+        assert_eq!(registry.len(), 1);
+    }
+
+    #[test]
+    fn rejects_missing_team_operations() {
+        let registry = TeamRegistry::new();
+        assert!(registry.delete("nonexistent").is_err());
+        assert!(registry.get("nonexistent").is_none());
+    }
+
+    // ── Cron tests ──────────────────────────────────────
+
+    #[test]
+    fn creates_and_retrieves_cron() {
+        let registry = CronRegistry::new();
+        let entry = registry.create("0 * * * *", "Check status", Some("hourly check"));
+        assert_eq!(entry.schedule, "0 * * * *");
+        assert_eq!(entry.prompt, "Check status");
+        assert!(entry.enabled);
+        assert_eq!(entry.run_count, 0);
+        assert!(entry.last_run_at.is_none());
+
+        let fetched = registry.get(&entry.cron_id).expect("cron should exist");
+        assert_eq!(fetched.cron_id, entry.cron_id);
+    }
+
+    #[test]
+    fn lists_with_enabled_filter() {
+        let registry = CronRegistry::new();
+        let c1 = registry.create("* * * * *", "Task 1", None);
+        let c2 = registry.create("0 * * * *", "Task 2", None);
+        registry
+            .disable(&c1.cron_id)
+            .expect("disable should succeed");
+
+        let all = registry.list(false);
+        assert_eq!(all.len(), 2);
+
+        let enabled_only = registry.list(true);
+        assert_eq!(enabled_only.len(), 1);
+        assert_eq!(enabled_only[0].cron_id, c2.cron_id);
+    }
+
+    #[test]
+    fn deletes_cron_entry() {
+        let registry = CronRegistry::new();
+        let entry = registry.create("* * * * *", "To delete", None);
+        let deleted = registry
+            .delete(&entry.cron_id)
+            .expect("delete should succeed");
+        assert_eq!(deleted.cron_id, entry.cron_id);
+        assert!(registry.get(&entry.cron_id).is_none());
+        assert!(registry.is_empty());
+    }
+
+    #[test]
+    fn records_cron_runs() {
+        let registry = CronRegistry::new();
+        let entry = registry.create("*/5 * * * *", "Recurring", None);
+        registry.record_run(&entry.cron_id).unwrap();
+        registry.record_run(&entry.cron_id).unwrap();
+
+        let fetched = registry.get(&entry.cron_id).unwrap();
+        assert_eq!(fetched.run_count, 2);
+        assert!(fetched.last_run_at.is_some());
+    }
+
+    #[test]
+    fn rejects_missing_cron_operations() {
+        let registry = CronRegistry::new();
+        assert!(registry.delete("nonexistent").is_err());
+        assert!(registry.disable("nonexistent").is_err());
+        assert!(registry.record_run("nonexistent").is_err());
+        assert!(registry.get("nonexistent").is_none());
+    }
+
+    #[test]
+    fn team_status_display_all_variants() {
+        // given
+        let cases = [
+            (TeamStatus::Created, "created"),
+            (TeamStatus::Running, "running"),
+            (TeamStatus::Completed, "completed"),
+            (TeamStatus::Deleted, "deleted"),
+        ];
+
+        // when
+        let rendered: Vec<_> = cases
+            .into_iter()
+            .map(|(status, expected)| (status.to_string(), expected))
+            .collect();
+
+        // then
+        assert_eq!(
+            rendered,
+            vec![
+                ("created".to_string(), "created"),
+                ("running".to_string(), "running"),
+                ("completed".to_string(), "completed"),
+                ("deleted".to_string(), "deleted"),
+            ]
+        );
+    }
+
+    #[test]
+    fn new_team_registry_is_empty() {
+        // given
+        let registry = TeamRegistry::new();
+
+        // when
+        let teams = registry.list();
+
+        // then
+        assert!(registry.is_empty());
+        assert_eq!(registry.len(), 0);
+        assert!(teams.is_empty());
+    }
+
+    #[test]
+    fn team_remove_nonexistent_returns_none() {
+        // given
+        let registry = TeamRegistry::new();
+
+        // when
+        let removed = registry.remove("missing");
+
+        // then
+        assert!(removed.is_none());
+    }
+
+    #[test]
+    fn team_len_transitions() {
+        // given
+        let registry = TeamRegistry::new();
+
+        // when
+        let alpha = registry.create("Alpha", vec![]);
+        let beta = registry.create("Beta", vec![]);
+        let after_create = registry.len();
+        registry.remove(&alpha.team_id);
+        let after_first_remove = registry.len();
+        registry.remove(&beta.team_id);
+
+        // then
+        assert_eq!(after_create, 2);
+        assert_eq!(after_first_remove, 1);
+        assert_eq!(registry.len(), 0);
+        assert!(registry.is_empty());
+    }
+
+    #[test]
+    fn cron_list_all_disabled_returns_empty_for_enabled_only() {
+        // given
+        let registry = CronRegistry::new();
+        let first = registry.create("* * * * *", "Task 1", None);
+        let second = registry.create("0 * * * *", "Task 2", None);
+        registry
+            .disable(&first.cron_id)
+            .expect("disable should succeed");
+        registry
+            .disable(&second.cron_id)
+            .expect("disable should succeed");
+
+        // when
+        let enabled_only = registry.list(true);
+        let all_entries = registry.list(false);
+
+        // then
+        assert!(enabled_only.is_empty());
+        assert_eq!(all_entries.len(), 2);
+    }
+
+    #[test]
+    fn cron_create_without_description() {
+        // given
+        let registry = CronRegistry::new();
+
+        // when
+        let entry = registry.create("*/15 * * * *", "Check health", None);
+
+        // then
+        assert!(entry.cron_id.starts_with("cron_"));
+        assert_eq!(entry.description, None);
+        assert!(entry.enabled);
+        assert_eq!(entry.run_count, 0);
+        assert_eq!(entry.last_run_at, None);
+    }
+
+    #[test]
+    fn new_cron_registry_is_empty() {
+        // given
+        let registry = CronRegistry::new();
+
+        // when
+        let enabled_only = registry.list(true);
+        let all_entries = registry.list(false);
+
+        // then
+        assert!(registry.is_empty());
+        assert_eq!(registry.len(), 0);
+        assert!(enabled_only.is_empty());
+        assert!(all_entries.is_empty());
+    }
+
+    #[test]
+    fn cron_record_run_updates_timestamp_and_counter() {
+        // given
+        let registry = CronRegistry::new();
+        let entry = registry.create("*/5 * * * *", "Recurring", None);
+
+        // when
+        registry
+            .record_run(&entry.cron_id)
+            .expect("first run should succeed");
+        registry
+            .record_run(&entry.cron_id)
+            .expect("second run should succeed");
+        let fetched = registry.get(&entry.cron_id).expect("entry should exist");
+
+        // then
+        assert_eq!(fetched.run_count, 2);
+        assert!(fetched.last_run_at.is_some());
+        assert!(fetched.updated_at >= entry.updated_at);
+    }
+
+    #[test]
+    fn cron_disable_updates_timestamp() {
+        // given
+        let registry = CronRegistry::new();
+        let entry = registry.create("0 0 * * *", "Nightly", None);
+
+        // when
+        registry
+            .disable(&entry.cron_id)
+            .expect("disable should succeed");
+        let fetched = registry.get(&entry.cron_id).expect("entry should exist");
+
+        // then
+        assert!(!fetched.enabled);
+        assert!(fetched.updated_at >= entry.updated_at);
+    }
+}
@@ -0,0 +1,299 @@
+use std::path::{Path, PathBuf};
+
+const TRUST_PROMPT_CUES: &[&str] = &[
+    "do you trust the files in this folder",
+    "trust the files in this folder",
+    "trust this folder",
+    "allow and continue",
+    "yes, proceed",
+];
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum TrustPolicy {
+    AutoTrust,
+    RequireApproval,
+    Deny,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum TrustEvent {
+    TrustRequired { cwd: String },
+    TrustResolved { cwd: String, policy: TrustPolicy },
+    TrustDenied { cwd: String, reason: String },
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct TrustConfig {
+    allowlisted: Vec<PathBuf>,
+    denied: Vec<PathBuf>,
+}
+
+impl TrustConfig {
+    #[must_use]
+    pub fn new() -> Self {
+        Self::default()
+    }
+
+    #[must_use]
+    pub fn with_allowlisted(mut self, path: impl Into<PathBuf>) -> Self {
+        self.allowlisted.push(path.into());
+        self
+    }
+
+    #[must_use]
+    pub fn with_denied(mut self, path: impl Into<PathBuf>) -> Self {
+        self.denied.push(path.into());
+        self
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum TrustDecision {
+    NotRequired,
+    Required {
+        policy: TrustPolicy,
+        events: Vec<TrustEvent>,
+    },
+}
+
+impl TrustDecision {
+    #[must_use]
+    pub fn policy(&self) -> Option<TrustPolicy> {
+        match self {
+            Self::NotRequired => None,
+            Self::Required { policy, .. } => Some(*policy),
+        }
+    }
+
+    #[must_use]
+    pub fn events(&self) -> &[TrustEvent] {
+        match self {
+            Self::NotRequired => &[],
+            Self::Required { events, .. } => events,
+        }
+    }
+}
+
+#[derive(Debug, Clone)]
+pub struct TrustResolver {
+    config: TrustConfig,
+}
+
+impl TrustResolver {
+    #[must_use]
+    pub fn new(config: TrustConfig) -> Self {
+        Self { config }
+    }
+
+    #[must_use]
+    pub fn resolve(&self, cwd: &str, screen_text: &str) -> TrustDecision {
+        if !detect_trust_prompt(screen_text) {
+            return TrustDecision::NotRequired;
+        }
+
+        let mut events = vec![TrustEvent::TrustRequired {
+            cwd: cwd.to_owned(),
+        }];
+
+        if let Some(matched_root) = self
+            .config
+            .denied
+            .iter()
+            .find(|root| path_matches(cwd, root))
+        {
+            let reason = format!("cwd matches denied trust root: {}", matched_root.display());
+            events.push(TrustEvent::TrustDenied {
+                cwd: cwd.to_owned(),
+                reason,
+            });
+            return TrustDecision::Required {
+                policy: TrustPolicy::Deny,
+                events,
+            };
+        }
+
+        if self
+            .config
+            .allowlisted
+            .iter()
+            .any(|root| path_matches(cwd, root))
+        {
+            events.push(TrustEvent::TrustResolved {
+                cwd: cwd.to_owned(),
+                policy: TrustPolicy::AutoTrust,
+            });
+            return TrustDecision::Required {
+                policy: TrustPolicy::AutoTrust,
+                events,
+            };
+        }
+
+        TrustDecision::Required {
+            policy: TrustPolicy::RequireApproval,
+            events,
+        }
+    }
+
+    #[must_use]
+    pub fn trusts(&self, cwd: &str) -> bool {
+        !self
+            .config
+            .denied
+            .iter()
+            .any(|root| path_matches(cwd, root))
+            && self
+                .config
+                .allowlisted
+                .iter()
+                .any(|root| path_matches(cwd, root))
+    }
+}
+
+#[must_use]
+pub fn detect_trust_prompt(screen_text: &str) -> bool {
+    let lowered = screen_text.to_ascii_lowercase();
+    TRUST_PROMPT_CUES
+        .iter()
+        .any(|needle| lowered.contains(needle))
+}
+
+#[must_use]
+pub fn path_matches_trusted_root(cwd: &str, trusted_root: &str) -> bool {
+    path_matches(cwd, &normalize_path(Path::new(trusted_root)))
+}
+
+fn path_matches(candidate: &str, root: &Path) -> bool {
+    let candidate = normalize_path(Path::new(candidate));
+    let root = normalize_path(root);
+    candidate == root || candidate.starts_with(&root)
+}
+
+fn normalize_path(path: &Path) -> PathBuf {
+    std::fs::canonicalize(path).unwrap_or_else(|_| path.to_path_buf())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{
+        detect_trust_prompt, path_matches_trusted_root, TrustConfig, TrustDecision, TrustEvent,
+        TrustPolicy, TrustResolver,
+    };
+
+    #[test]
+    fn detects_known_trust_prompt_copy() {
+        // given
+        let screen_text = "Do you trust the files in this folder?\n1. Yes, proceed\n2. No";
+
+        // when
+        let detected = detect_trust_prompt(screen_text);
+
+        // then
+        assert!(detected);
+    }
+
+    #[test]
+    fn does_not_emit_events_when_prompt_is_absent() {
+        // given
+        let resolver = TrustResolver::new(TrustConfig::new().with_allowlisted("/tmp/worktrees"));
+
+        // when
+        let decision = resolver.resolve("/tmp/worktrees/repo-a", "Ready for your input\n>");
+
+        // then
+        assert_eq!(decision, TrustDecision::NotRequired);
+        assert_eq!(decision.events(), &[]);
+        assert_eq!(decision.policy(), None);
+    }
+
+    #[test]
+    fn auto_trusts_allowlisted_cwd_after_prompt_detection() {
+        // given
+        let resolver = TrustResolver::new(TrustConfig::new().with_allowlisted("/tmp/worktrees"));
+
+        // when
+        let decision = resolver.resolve(
+            "/tmp/worktrees/repo-a",
+            "Do you trust the files in this folder?\n1. Yes, proceed\n2. No",
+        );
+
+        // then
+        assert_eq!(decision.policy(), Some(TrustPolicy::AutoTrust));
+        assert_eq!(
+            decision.events(),
+            &[
+                TrustEvent::TrustRequired {
+                    cwd: "/tmp/worktrees/repo-a".to_string(),
+                },
+                TrustEvent::TrustResolved {
+                    cwd: "/tmp/worktrees/repo-a".to_string(),
+                    policy: TrustPolicy::AutoTrust,
+                },
+            ]
+        );
+    }
+
+    #[test]
+    fn requires_approval_for_unknown_cwd_after_prompt_detection() {
+        // given
+        let resolver = TrustResolver::new(TrustConfig::new().with_allowlisted("/tmp/worktrees"));
+
+        // when
+        let decision = resolver.resolve(
+            "/tmp/other/repo-b",
+            "Do you trust the files in this folder?\n1. Yes, proceed\n2. No",
+        );
+
+        // then
+        assert_eq!(decision.policy(), Some(TrustPolicy::RequireApproval));
+        assert_eq!(
+            decision.events(),
+            &[TrustEvent::TrustRequired {
+                cwd: "/tmp/other/repo-b".to_string(),
+            }]
+        );
+    }
+
+    #[test]
+    fn denied_root_takes_precedence_over_allowlist() {
+        // given
+        let resolver = TrustResolver::new(
+            TrustConfig::new()
+                .with_allowlisted("/tmp/worktrees")
+                .with_denied("/tmp/worktrees/repo-c"),
+        );
+
+        // when
+        let decision = resolver.resolve(
+            "/tmp/worktrees/repo-c",
+            "Do you trust the files in this folder?\n1. Yes, proceed\n2. No",
+        );
+
+        // then
+        assert_eq!(decision.policy(), Some(TrustPolicy::Deny));
+        assert_eq!(
+            decision.events(),
+            &[
+                TrustEvent::TrustRequired {
+                    cwd: "/tmp/worktrees/repo-c".to_string(),
+                },
+                TrustEvent::TrustDenied {
+                    cwd: "/tmp/worktrees/repo-c".to_string(),
+                    reason: "cwd matches denied trust root: /tmp/worktrees/repo-c".to_string(),
+                },
+            ]
+        );
+    }
+
+    #[test]
+    fn sibling_prefix_does_not_match_trusted_root() {
+        // given
+        let trusted_root = "/tmp/worktrees";
+        let sibling_path = "/tmp/worktrees-other/repo-d";
+
+        // when
+        let matched = path_matches_trusted_root(sibling_path, trusted_root);
+
+        // then
+        assert!(!matched);
+    }
+}
@@ -1,11 +1,11 @@
 use crate::session::Session;
-use serde::{Deserialize, Serialize};

 const DEFAULT_INPUT_COST_PER_MILLION: f64 = 15.0;
 const DEFAULT_OUTPUT_COST_PER_MILLION: f64 = 75.0;
 const DEFAULT_CACHE_CREATION_COST_PER_MILLION: f64 = 18.75;
 const DEFAULT_CACHE_READ_COST_PER_MILLION: f64 = 1.5;

+/// Per-million-token pricing used for cost estimation.
 #[derive(Debug, Clone, Copy, PartialEq)]
 pub struct ModelPricing {
    pub input_cost_per_million: f64,
@@ -26,7 +26,8 @@ impl ModelPricing {
    }
 }

-#[derive(Debug, Clone, Copy, Serialize, Deserialize, Default, PartialEq, Eq)]
+/// Token counters accumulated for a conversation turn or session.
+#[derive(Debug, Clone, Copy, Default, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
 pub struct TokenUsage {
    pub input_tokens: u32,
    pub output_tokens: u32,
@@ -34,6 +35,7 @@ pub struct TokenUsage {
    pub cache_read_input_tokens: u32,
 }

+/// Estimated dollar cost derived from a [`TokenUsage`] sample.
 #[derive(Debug, Clone, Copy, PartialEq)]
 pub struct UsageCostEstimate {
    pub input_cost_usd: f64,
@@ -52,6 +54,7 @@ impl UsageCostEstimate {
    }
 }

+/// Returns pricing metadata for a known model alias or family.
 #[must_use]
 pub fn pricing_for_model(model: &str) -> Option<ModelPricing> {
    let normalized = model.to_ascii_lowercase();
@@ -156,10 +159,12 @@ fn cost_for_tokens(tokens: u32, usd_per_million_tokens: f64) -> f64 {
 }

 #[must_use]
+/// Formats a dollar-denominated value for CLI display.
 pub fn format_usd(amount: f64) -> String {
    format!("${amount:.4}")
 }

+/// Aggregates token usage across a running session.
 #[derive(Debug, Clone, Default, PartialEq, Eq)]
 pub struct UsageTracker {
    latest_turn: TokenUsage,
@@ -250,9 +255,9 @@ mod tests {
        let cost = usage.estimate_cost_usd();
        assert_eq!(format_usd(cost.input_cost_usd), "$15.0000");
        assert_eq!(format_usd(cost.output_cost_usd), "$37.5000");
-        let lines = usage.summary_lines_for_model("usage", Some("claude-sonnet-4-6"));
+        let lines = usage.summary_lines_for_model("usage", Some("claude-sonnet-4-20250514"));
        assert!(lines[0].contains("estimated_cost=$54.6750"));
-        assert!(lines[0].contains("model=claude-sonnet-4-6"));
+        assert!(lines[0].contains("model=claude-sonnet-4-20250514"));
        assert!(lines[1].contains("cache_read=$0.3000"));
    }

@@ -265,7 +270,7 @@ mod tests {
            cache_read_input_tokens: 0,
        };

-        let haiku = pricing_for_model("claude-haiku-4-5-20251213").expect("haiku pricing");
+        let haiku = pricing_for_model("claude-haiku-4-5-20251001").expect("haiku pricing");
        let opus = pricing_for_model("claude-opus-4-6").expect("opus pricing");
        let haiku_cost = usage.estimate_cost_usd_with_pricing(haiku);
        let opus_cost = usage.estimate_cost_usd_with_pricing(opus);
@@ -287,21 +292,19 @@ mod tests {

    #[test]
    fn reconstructs_usage_from_session_messages() {
-        let session = Session {
-            version: 1,
-            messages: vec![ConversationMessage {
-                role: MessageRole::Assistant,
-                blocks: vec![ContentBlock::Text {
-                    text: "done".to_string(),
-                }],
-                usage: Some(TokenUsage {
-                    input_tokens: 5,
-                    output_tokens: 2,
-                    cache_creation_input_tokens: 1,
-                    cache_read_input_tokens: 0,
-                }),
+        let mut session = Session::new();
+        session.messages = vec![ConversationMessage {
+            role: MessageRole::Assistant,
+            blocks: vec![ContentBlock::Text {
+                text: "done".to_string(),
            }],
-        };
+            usage: Some(TokenUsage {
+                input_tokens: 5,
+                output_tokens: 2,
+                cache_creation_input_tokens: 1,
+                cache_read_input_tokens: 0,
+            }),
+        }];

        let tracker = UsageTracker::from_session(&session);
        assert_eq!(tracker.turns(), 1);
@@ -0,0 +1,386 @@
+#![allow(clippy::doc_markdown, clippy::uninlined_format_args, unused_imports)]
+//! Integration tests for cross-module wiring.
+//!
+//! These tests verify that adjacent modules in the runtime crate actually
+//! connect correctly — catching wiring gaps that unit tests miss.
+
+use std::time::Duration;
+
+use runtime::green_contract::{GreenContract, GreenContractOutcome, GreenLevel};
+use runtime::{
+    apply_policy, BranchFreshness, DiffScope, LaneBlocker, LaneContext, PolicyAction,
+    PolicyCondition, PolicyEngine, PolicyRule, ReconcileReason, ReviewStatus, StaleBranchAction,
+    StaleBranchPolicy,
+};
+
+/// stale_branch + policy_engine integration:
+/// When a branch is detected stale, does it correctly flow through
+/// PolicyCondition::StaleBranch to generate the expected action?
+#[test]
+fn stale_branch_detection_flows_into_policy_engine() {
+    // given — a stale branch context (2 hours behind main, threshold is 1 hour)
+    let stale_context = LaneContext::new(
+        "stale-lane",
+        0,
+        Duration::from_secs(2 * 60 * 60), // 2 hours stale
+        LaneBlocker::None,
+        ReviewStatus::Pending,
+        DiffScope::Full,
+        false,
+    );
+
+    let engine = PolicyEngine::new(vec![PolicyRule::new(
+        "stale-merge-forward",
+        PolicyCondition::StaleBranch,
+        PolicyAction::MergeForward,
+        10,
+    )]);
+
+    // when
+    let actions = engine.evaluate(&stale_context);
+
+    // then
+    assert_eq!(actions, vec![PolicyAction::MergeForward]);
+}
+
+/// stale_branch + policy_engine: Fresh branch does NOT trigger stale rules
+#[test]
+fn fresh_branch_does_not_trigger_stale_policy() {
+    let fresh_context = LaneContext::new(
+        "fresh-lane",
+        0,
+        Duration::from_secs(30 * 60), // 30 min stale — under 1 hour threshold
+        LaneBlocker::None,
+        ReviewStatus::Pending,
+        DiffScope::Full,
+        false,
+    );
+
+    let engine = PolicyEngine::new(vec![PolicyRule::new(
+        "stale-merge-forward",
+        PolicyCondition::StaleBranch,
+        PolicyAction::MergeForward,
+        10,
+    )]);
+
+    let actions = engine.evaluate(&fresh_context);
+    assert!(actions.is_empty());
+}
+
+/// green_contract + policy_engine integration:
+/// A lane that meets its green contract should be mergeable
+#[test]
+fn green_contract_satisfied_allows_merge() {
+    let contract = GreenContract::new(GreenLevel::Workspace);
+    let satisfied = contract.is_satisfied_by(GreenLevel::Workspace);
+    assert!(satisfied);
+
+    let exceeded = contract.is_satisfied_by(GreenLevel::MergeReady);
+    assert!(exceeded);
+
+    let insufficient = contract.is_satisfied_by(GreenLevel::Package);
+    assert!(!insufficient);
+}
+
+/// green_contract + policy_engine:
+/// Lane with green level below contract requirement gets blocked
+#[test]
+fn green_contract_unsatisfied_blocks_merge() {
+    let context = LaneContext::new(
+        "partial-green-lane",
+        1, // GreenLevel::Package as u8
+        Duration::from_secs(0),
+        LaneBlocker::None,
+        ReviewStatus::Pending,
+        DiffScope::Full,
+        false,
+    );
+
+    // This is a conceptual test — we need a way to express "requires workspace green"
+    // Currently LaneContext has raw green_level: u8, not a contract
+    // For now we just verify the policy condition works
+    let engine = PolicyEngine::new(vec![PolicyRule::new(
+        "workspace-green-required",
+        PolicyCondition::GreenAt { level: 3 }, // GreenLevel::Workspace
+        PolicyAction::MergeToDev,
+        10,
+    )]);
+
+    let actions = engine.evaluate(&context);
+    assert!(actions.is_empty()); // level 1 < 3, so no merge
+}
+
+/// reconciliation + policy_engine integration:
+/// A reconciled lane should be handled by reconcile rules, not generic closeout
+#[test]
+fn reconciled_lane_matches_reconcile_condition() {
+    let context = LaneContext::reconciled("reconciled-lane");
+
+    let engine = PolicyEngine::new(vec![
+        PolicyRule::new(
+            "reconcile-first",
+            PolicyCondition::LaneReconciled,
+            PolicyAction::Reconcile {
+                reason: ReconcileReason::AlreadyMerged,
+            },
+            5,
+        ),
+        PolicyRule::new(
+            "generic-closeout",
+            PolicyCondition::LaneCompleted,
+            PolicyAction::CloseoutLane,
+            30,
+        ),
+    ]);
+
+    let actions = engine.evaluate(&context);
+
+    // Both rules fire — reconcile (priority 5) first, then closeout (priority 30)
+    assert_eq!(
+        actions,
+        vec![
+            PolicyAction::Reconcile {
+                reason: ReconcileReason::AlreadyMerged,
+            },
+            PolicyAction::CloseoutLane,
+        ]
+    );
+}
+
+/// stale_branch module: apply_policy generates correct actions
+#[test]
+fn stale_branch_apply_policy_produces_rebase_action() {
+    let stale = BranchFreshness::Stale {
+        commits_behind: 5,
+        missing_fixes: vec!["fix-123".to_string()],
+    };
+
+    let action = apply_policy(&stale, StaleBranchPolicy::AutoRebase);
+    assert_eq!(action, StaleBranchAction::Rebase);
+}
+
+#[test]
+fn stale_branch_apply_policy_produces_merge_forward_action() {
+    let stale = BranchFreshness::Stale {
+        commits_behind: 3,
+        missing_fixes: vec![],
+    };
+
+    let action = apply_policy(&stale, StaleBranchPolicy::AutoMergeForward);
+    assert_eq!(action, StaleBranchAction::MergeForward);
+}
+
+#[test]
+fn stale_branch_apply_policy_warn_only() {
+    let stale = BranchFreshness::Stale {
+        commits_behind: 2,
+        missing_fixes: vec!["fix-456".to_string()],
+    };
+
+    let action = apply_policy(&stale, StaleBranchPolicy::WarnOnly);
+    match action {
+        StaleBranchAction::Warn { message } => {
+            assert!(message.contains("2 commit(s) behind main"));
+            assert!(message.contains("fix-456"));
+        }
+        _ => panic!("expected Warn action, got {:?}", action),
+    }
+}
+
+#[test]
+fn stale_branch_fresh_produces_noop() {
+    let fresh = BranchFreshness::Fresh;
+    let action = apply_policy(&fresh, StaleBranchPolicy::AutoRebase);
+    assert_eq!(action, StaleBranchAction::Noop);
+}
+
+/// Combined flow: stale detection + policy + action
+#[test]
+fn end_to_end_stale_lane_gets_merge_forward_action() {
+    // Simulating what a harness would do:
+    // 1. Detect branch freshness
+    // 2. Build lane context from freshness + other signals
+    // 3. Run policy engine
+    // 4. Return actions
+
+    // given: detected stale state
+    let _freshness = BranchFreshness::Stale {
+        commits_behind: 5,
+        missing_fixes: vec!["fix-123".to_string()],
+    };
+
+    // when: build context and evaluate policy
+    let context = LaneContext::new(
+        "lane-9411",
+        3,                                // Workspace green
+        Duration::from_secs(5 * 60 * 60), // 5 hours stale, definitely over threshold
+        LaneBlocker::None,
+        ReviewStatus::Approved,
+        DiffScope::Scoped,
+        false,
+    );
+
+    let engine = PolicyEngine::new(vec![
+        // Priority 5: Check if stale first
+        PolicyRule::new(
+            "auto-merge-forward-if-stale-and-approved",
+            PolicyCondition::And(vec![
+                PolicyCondition::StaleBranch,
+                PolicyCondition::ReviewPassed,
+            ]),
+            PolicyAction::MergeForward,
+            5,
+        ),
+        // Priority 10: Normal stale handling
+        PolicyRule::new(
+            "stale-warning",
+            PolicyCondition::StaleBranch,
+            PolicyAction::Notify {
+                channel: "#build-status".to_string(),
+            },
+            10,
+        ),
+    ]);
+
+    let actions = engine.evaluate(&context);
+
+    // then: both rules should fire (stale + approved matches both)
+    assert_eq!(
+        actions,
+        vec![
+            PolicyAction::MergeForward,
+            PolicyAction::Notify {
+                channel: "#build-status".to_string(),
+            },
+        ]
+    );
+}
+
+/// Fresh branch with approved review should merge (not stale-blocked)
+#[test]
+fn fresh_approved_lane_gets_merge_action() {
+    let context = LaneContext::new(
+        "fresh-approved-lane",
+        3,                            // Workspace green
+        Duration::from_secs(30 * 60), // 30 min — under 1 hour threshold = fresh
+        LaneBlocker::None,
+        ReviewStatus::Approved,
+        DiffScope::Scoped,
+        false,
+    );
+
+    let engine = PolicyEngine::new(vec![PolicyRule::new(
+        "merge-if-green-approved-not-stale",
+        PolicyCondition::And(vec![
+            PolicyCondition::GreenAt { level: 3 },
+            PolicyCondition::ReviewPassed,
+            // NOT PolicyCondition::StaleBranch — fresh lanes bypass this
+        ]),
+        PolicyAction::MergeToDev,
+        5,
+    )]);
+
+    let actions = engine.evaluate(&context);
+    assert_eq!(actions, vec![PolicyAction::MergeToDev]);
+}
+
+/// worker_boot + recovery_recipes + policy_engine integration:
+/// When a session completes with a provider failure, does the worker
+/// status transition trigger the correct recovery recipe, and does
+/// the resulting recovery state feed into policy decisions?
+#[test]
+fn worker_provider_failure_flows_through_recovery_to_policy() {
+    use runtime::recovery_recipes::{
+        attempt_recovery, FailureScenario, RecoveryContext, RecoveryResult, RecoveryStep,
+    };
+    use runtime::worker_boot::{WorkerFailureKind, WorkerRegistry, WorkerStatus};
+
+    // given — a worker that encounters a provider failure during session completion
+    let registry = WorkerRegistry::new();
+    let worker = registry.create("/tmp/repo-recovery-test", &[], true);
+
+    // Worker reaches ready state
+    registry
+        .observe(&worker.worker_id, "Ready for your input\n>")
+        .expect("ready observe should succeed");
+    registry
+        .send_prompt(&worker.worker_id, Some("Run analysis"))
+        .expect("prompt send should succeed");
+
+    // Session completes with provider failure (finish="unknown", tokens=0)
+    let failed_worker = registry
+        .observe_completion(&worker.worker_id, "unknown", 0)
+        .expect("completion observe should succeed");
+    assert_eq!(failed_worker.status, WorkerStatus::Failed);
+    let failure = failed_worker
+        .last_error
+        .expect("worker should have recorded error");
+    assert_eq!(failure.kind, WorkerFailureKind::Provider);
+
+    // Bridge: WorkerFailureKind -> FailureScenario
+    let scenario = FailureScenario::from_worker_failure_kind(failure.kind);
+    assert_eq!(scenario, FailureScenario::ProviderFailure);
+
+    // Recovery recipe lookup and execution
+    let mut ctx = RecoveryContext::new();
+    let result = attempt_recovery(&scenario, &mut ctx);
+
+    // then — recovery should recommend RestartWorker step
+    assert!(
+        matches!(result, RecoveryResult::Recovered { steps_taken: 1 }),
+        "provider failure should recover via single RestartWorker step, got: {result:?}"
+    );
+    assert!(
+        ctx.events().iter().any(|e| {
+            matches!(
+                e,
+                runtime::recovery_recipes::RecoveryEvent::RecoveryAttempted {
+                    result: RecoveryResult::Recovered { steps_taken: 1 },
+                    ..
+                }
+            )
+        }),
+        "recovery should emit structured attempt event"
+    );
+
+    // Policy integration: recovery success + green status = merge-ready
+    // (Simulating the policy check that would happen after successful recovery)
+    let recovery_success = matches!(result, RecoveryResult::Recovered { .. });
+    let green_level = 3; // Workspace green
+    let not_stale = Duration::from_secs(30 * 60); // 30 min — fresh
+
+    let post_recovery_context = LaneContext::new(
+        "recovered-lane",
+        green_level,
+        not_stale,
+        LaneBlocker::None,
+        ReviewStatus::Approved,
+        DiffScope::Scoped,
+        false,
+    );
+
+    let policy_engine = PolicyEngine::new(vec![
+        // Rule: if recovered from failure + green + approved -> merge
+        PolicyRule::new(
+            "merge-after-successful-recovery",
+            PolicyCondition::And(vec![
+                PolicyCondition::GreenAt { level: 3 },
+                PolicyCondition::ReviewPassed,
+            ]),
+            PolicyAction::MergeToDev,
+            10,
+        ),
+    ]);
+
+    // Recovery success is a pre-condition; policy evaluates post-recovery context
+    assert!(
+        recovery_success,
+        "recovery must succeed for lane to proceed"
+    );
+    let actions = policy_engine.evaluate(&post_recovery_context);
+    assert_eq!(
+        actions,
+        vec![PolicyAction::MergeToDev],
+        "post-recovery green+approved lane should be merge-ready"
+    );
+}
@@ -15,12 +15,20 @@ commands = { path = "../commands" }
 compat-harness = { path = "../compat-harness" }
 crossterm = "0.28"
 pulldown-cmark = "0.13"
-plugins = { path = "../plugins" }
+rustyline = "15"
 runtime = { path = "../runtime" }
-serde_json = "1"
+plugins = { path = "../plugins" }
+serde = { version = "1", features = ["derive"] }
+serde_json.workspace = true
 syntect = "5"
-tokio = { version = "1", features = ["rt-multi-thread", "time"] }
+tokio = { version = "1", features = ["rt-multi-thread", "signal", "time"] }
 tools = { path = "../tools" }

 [lints]
 workspace = true
+
+[dev-dependencies]
+mock-anthropic-service = { path = "../mock-anthropic-service" }
+serde_json.workspace = true
+tokio = { version = "1", features = ["rt-multi-thread"] }
+
@@ -0,0 +1,64 @@
+# Rusty Claude CLI 模块 (rusty-claude-cli)
+
+本模块提供了 Claw 命令行界面的另一个功能完整的实现。它集成了对话、工具执行、插件扩展以及身份验证等核心功能。
+
+## 概览
+
+`rusty-claude-cli` 是一个全功能的 CLI 应用程序，其主要职责包括：
+- **用户交互**：提供交互式 REPL 和非交互式命令执行（`prompt` 子命令）。
+- **环境初始化**：处理项目初始化 (`init`) 和配置加载。
+- **身份验证**：通过本地回环服务器处理 OAuth 登录流程。
+- **状态渲染**：实现丰富的终端 UI 效果，如 Markdown 渲染、语法高亮和动态加载动画 (Spinner)。
+- **会话管理**：支持从保存的文件中恢复会话并执行追加的斜杠命令。
+
+## 与 `claw-cli` 的关系
+
+虽然 `rusty-claude-cli` 和 `claw-cli` 都生成名为 `claw` 的二进制文件，但 `rusty-claude-cli` 包含更复杂的集成逻辑：
+- 它直接引用了几乎所有的核心 crate（`runtime`, `api`, `tools`, `plugins`, `commands`）。
+- 它的 `main.rs` 实现非常庞大，包含了大量的业务编排逻辑。
+- 它可以作为一个独立的、集成度极高的 CLI 参考实现。
+
+## 关键特性
+
+- **多功能子命令**：
+  - `prompt`：快速运行单次推理。
+  - `login`/`logout`：OAuth 身份验证管理。
+  - `init`：项目环境自举。
+  - `bootstrap-plan`：查看系统的启动阶段。
+  - `dump-manifests`：从上游源码中提取并显示功能清单。
+- **增强的 REPL**：
+  - 支持多行输入和历史记录。
+  - 集成了斜杠命令处理引擎。
+  - 提供详细的消耗统计和权限模式切换报告。
+- **灵活的权限控制**：支持通过命令行参数 `--permission-mode` 或环境变量动态调整权限级别。
+
+## 实现逻辑
+
+### 核心子模块
+
+- **`main.rs`**: 程序的入口，包含了复杂的参数解析逻辑和 REPL 循环。
+- **`render.rs`**: 封装了 `TerminalRenderer` 和 `Spinner`，负责所有的终端输出美化。
+- **`input.rs`**: 处理从标准输入读取数据及命令解析。
+- **`init.rs`**: 专注于仓库的初始化和 `.claw.md` 文件的生成。
+- **`app.rs`**: 可能包含应用程序级别的高层状态管理（取决于具体实现）。
+
+### 工作流程
+
+1. 程序启动，解析命令行参数。
+2. 根据参数决定是执行单次任务还是进入 REPL 模式。
+3. 在 REPL 模式下，初始化 `ConversationRuntime`。
+4. 进入循环：读取用户输入 -> 处理斜杠命令或发送给 AI -> 渲染响应 -> 执行工具 -> 循环。
+5. 会话数据根据需要保存或恢复。
+
+## 使用示例
+
+```bash
+# 启动交互模式
+cargo run -p rusty-claude-cli --bin claw
+
+# 直接运行 Prompt
+cargo run -p rusty-claude-cli --bin claw prompt "检查代码中的内存泄漏"
+
+# 恢复之前的会话并执行压缩
+cargo run -p rusty-claude-cli --bin claw --resume session.json /compact
+```
@@ -0,0 +1,38 @@
+use std::env;
+use std::process::Command;
+
+fn main() {
+    // Get git SHA (short hash)
+    let git_sha = Command::new("git")
+        .args(["rev-parse", "--short", "HEAD"])
+        .output()
+        .ok()
+        .and_then(|output| {
+            if output.status.success() {
+                String::from_utf8(output.stdout).ok()
+            } else {
+                None
+            }
+        })
+        .map_or_else(|| "unknown".to_string(), |s| s.trim().to_string());
+
+    println!("cargo:rustc-env=GIT_SHA={git_sha}");
+
+    // TARGET is always set by Cargo during build
+    let target = env::var("TARGET").unwrap_or_else(|_| "unknown".to_string());
+    println!("cargo:rustc-env=TARGET={target}");
+
+    // Build date from SOURCE_DATE_EPOCH (reproducible builds) or current date.
+    let build_date = std::env::var("SOURCE_DATE_EPOCH")
+        .ok()
+        .and_then(|epoch| epoch.parse::<i64>().ok())
+        .map_or_else(
+            || "unknown".to_string(),
+            |_| std::env::var("BUILD_DATE").unwrap_or_else(|_| "unknown".to_string()),
+        );
+    println!("cargo:rustc-env=BUILD_DATE={build_date}");
+
+    // Rerun if git state changes
+    println!("cargo:rerun-if-changed=.git/HEAD");
+    println!("cargo:rerun-if-changed=.git/refs");
+}
@@ -1,398 +0,0 @@
-use std::io::{self, Write};
-use std::path::PathBuf;
-
-use crate::args::{OutputFormat, PermissionMode};
-use crate::input::{LineEditor, ReadOutcome};
-use crate::render::{Spinner, TerminalRenderer};
-use runtime::{ConversationClient, ConversationMessage, RuntimeError, StreamEvent, UsageSummary};
-
-#[derive(Debug, Clone, PartialEq, Eq)]
-pub struct SessionConfig {
-    pub model: String,
-    pub permission_mode: PermissionMode,
-    pub config: Option<PathBuf>,
-    pub output_format: OutputFormat,
-}
-
-#[derive(Debug, Clone, PartialEq, Eq)]
-pub struct SessionState {
-    pub turns: usize,
-    pub compacted_messages: usize,
-    pub last_model: String,
-    pub last_usage: UsageSummary,
-}
-
-impl SessionState {
-    #[must_use]
-    pub fn new(model: impl Into<String>) -> Self {
-        Self {
-            turns: 0,
-            compacted_messages: 0,
-            last_model: model.into(),
-            last_usage: UsageSummary::default(),
-        }
-    }
-}
-
-#[derive(Debug, Clone, Copy, PartialEq, Eq)]
-pub enum CommandResult {
-    Continue,
-}
-
-#[derive(Debug, Clone, PartialEq, Eq)]
-pub enum SlashCommand {
-    Help,
-    Status,
-    Compact,
-    Unknown(String),
-}
-
-impl SlashCommand {
-    #[must_use]
-    pub fn parse(input: &str) -> Option<Self> {
-        let trimmed = input.trim();
-        if !trimmed.starts_with('/') {
-            return None;
-        }
-
-        let command = trimmed
-            .trim_start_matches('/')
-            .split_whitespace()
-            .next()
-            .unwrap_or_default();
-        Some(match command {
-            "help" => Self::Help,
-            "status" => Self::Status,
-            "compact" => Self::Compact,
-            other => Self::Unknown(other.to_string()),
-        })
-    }
-}
-
-struct SlashCommandHandler {
-    command: SlashCommand,
-    summary: &'static str,
-}
-
-const SLASH_COMMAND_HANDLERS: &[SlashCommandHandler] = &[
-    SlashCommandHandler {
-        command: SlashCommand::Help,
-        summary: "Show command help",
-    },
-    SlashCommandHandler {
-        command: SlashCommand::Status,
-        summary: "Show current session status",
-    },
-    SlashCommandHandler {
-        command: SlashCommand::Compact,
-        summary: "Compact local session history",
-    },
-];
-
-pub struct CliApp {
-    config: SessionConfig,
-    renderer: TerminalRenderer,
-    state: SessionState,
-    conversation_client: ConversationClient,
-    conversation_history: Vec<ConversationMessage>,
-}
-
-impl CliApp {
-    pub fn new(config: SessionConfig) -> Result<Self, RuntimeError> {
-        let state = SessionState::new(config.model.clone());
-        let conversation_client = ConversationClient::from_env(config.model.clone())?;
-        Ok(Self {
-            config,
-            renderer: TerminalRenderer::new(),
-            state,
-            conversation_client,
-            conversation_history: Vec::new(),
-        })
-    }
-
-    pub fn run_repl(&mut self) -> io::Result<()> {
-        let mut editor = LineEditor::new("› ", Vec::new());
-        println!("Rusty Claude CLI interactive mode");
-        println!("Type /help for commands. Shift+Enter or Ctrl+J inserts a newline.");
-
-        loop {
-            match editor.read_line()? {
-                ReadOutcome::Submit(input) => {
-                    if input.trim().is_empty() {
-                        continue;
-                    }
-                    self.handle_submission(&input, &mut io::stdout())?;
-                }
-                ReadOutcome::Cancel => continue,
-                ReadOutcome::Exit => break,
-            }
-        }
-
-        Ok(())
-    }
-
-    pub fn run_prompt(&mut self, prompt: &str, out: &mut impl Write) -> io::Result<()> {
-        self.render_response(prompt, out)
-    }
-
-    pub fn handle_submission(
-        &mut self,
-        input: &str,
-        out: &mut impl Write,
-    ) -> io::Result<CommandResult> {
-        if let Some(command) = SlashCommand::parse(input) {
-            return self.dispatch_slash_command(command, out);
-        }
-
-        self.state.turns += 1;
-        self.render_response(input, out)?;
-        Ok(CommandResult::Continue)
-    }
-
-    fn dispatch_slash_command(
-        &mut self,
-        command: SlashCommand,
-        out: &mut impl Write,
-    ) -> io::Result<CommandResult> {
-        match command {
-            SlashCommand::Help => Self::handle_help(out),
-            SlashCommand::Status => self.handle_status(out),
-            SlashCommand::Compact => self.handle_compact(out),
-            SlashCommand::Unknown(name) => {
-                writeln!(out, "Unknown slash command: /{name}")?;
-                Ok(CommandResult::Continue)
-            }
-        }
-    }
-
-    fn handle_help(out: &mut impl Write) -> io::Result<CommandResult> {
-        writeln!(out, "Available commands:")?;
-        for handler in SLASH_COMMAND_HANDLERS {
-            let name = match handler.command {
-                SlashCommand::Help => "/help",
-                SlashCommand::Status => "/status",
-                SlashCommand::Compact => "/compact",
-                SlashCommand::Unknown(_) => continue,
-            };
-            writeln!(out, "  {name:<9} {}", handler.summary)?;
-        }
-        Ok(CommandResult::Continue)
-    }
-
-    fn handle_status(&mut self, out: &mut impl Write) -> io::Result<CommandResult> {
-        writeln!(
-            out,
-            "status: turns={} model={} permission-mode={:?} output-format={:?} last-usage={} in/{} out config={}",
-            self.state.turns,
-            self.state.last_model,
-            self.config.permission_mode,
-            self.config.output_format,
-            self.state.last_usage.input_tokens,
-            self.state.last_usage.output_tokens,
-            self.config
-                .config
-                .as_ref()
-                .map_or_else(|| String::from("<none>"), |path| path.display().to_string())
-        )?;
-        Ok(CommandResult::Continue)
-    }
-
-    fn handle_compact(&mut self, out: &mut impl Write) -> io::Result<CommandResult> {
-        self.state.compacted_messages += self.state.turns;
-        self.state.turns = 0;
-        self.conversation_history.clear();
-        writeln!(
-            out,
-            "Compacted session history into a local summary ({} messages total compacted).",
-            self.state.compacted_messages
-        )?;
-        Ok(CommandResult::Continue)
-    }
-
-    fn handle_stream_event(
-        renderer: &TerminalRenderer,
-        event: StreamEvent,
-        stream_spinner: &mut Spinner,
-        tool_spinner: &mut Spinner,
-        saw_text: &mut bool,
-        turn_usage: &mut UsageSummary,
-        out: &mut impl Write,
-    ) {
-        match event {
-            StreamEvent::TextDelta(delta) => {
-                if !*saw_text {
-                    let _ =
-                        stream_spinner.finish("Streaming response", renderer.color_theme(), out);
-                    *saw_text = true;
-                }
-                let _ = write!(out, "{delta}");
-                let _ = out.flush();
-            }
-            StreamEvent::ToolCallStart { name, input } => {
-                if *saw_text {
-                    let _ = writeln!(out);
-                }
-                let _ = tool_spinner.tick(
-                    &format!("Running tool `{name}` with {input}"),
-                    renderer.color_theme(),
-                    out,
-                );
-            }
-            StreamEvent::ToolCallResult {
-                name,
-                output,
-                is_error,
-            } => {
-                let label = if is_error {
-                    format!("Tool `{name}` failed")
-                } else {
-                    format!("Tool `{name}` completed")
-                };
-                let _ = tool_spinner.finish(&label, renderer.color_theme(), out);
-                let rendered_output = format!("### Tool `{name}`\n\n```text\n{output}\n```\n");
-                let _ = renderer.stream_markdown(&rendered_output, out);
-            }
-            StreamEvent::Usage(usage) => {
-                *turn_usage = usage;
-            }
-        }
-    }
-
-    fn write_turn_output(
-        &self,
-        summary: &runtime::TurnSummary,
-        out: &mut impl Write,
-    ) -> io::Result<()> {
-        match self.config.output_format {
-            OutputFormat::Text => {
-                writeln!(
-                    out,
-                    "\nToken usage: {} input / {} output",
-                    self.state.last_usage.input_tokens, self.state.last_usage.output_tokens
-                )?;
-            }
-            OutputFormat::Json => {
-                writeln!(
-                    out,
-                    "{}",
-                    serde_json::json!({
-                        "message": summary.assistant_text,
-                        "usage": {
-                            "input_tokens": self.state.last_usage.input_tokens,
-                            "output_tokens": self.state.last_usage.output_tokens,
-                        }
-                    })
-                )?;
-            }
-            OutputFormat::Ndjson => {
-                writeln!(
-                    out,
-                    "{}",
-                    serde_json::json!({
-                        "type": "message",
-                        "text": summary.assistant_text,
-                        "usage": {
-                            "input_tokens": self.state.last_usage.input_tokens,
-                            "output_tokens": self.state.last_usage.output_tokens,
-                        }
-                    })
-                )?;
-            }
-        }
-        Ok(())
-    }
-
-    fn render_response(&mut self, input: &str, out: &mut impl Write) -> io::Result<()> {
-        let mut stream_spinner = Spinner::new();
-        stream_spinner.tick(
-            "Opening conversation stream",
-            self.renderer.color_theme(),
-            out,
-        )?;
-
-        let mut turn_usage = UsageSummary::default();
-        let mut tool_spinner = Spinner::new();
-        let mut saw_text = false;
-        let renderer = &self.renderer;
-
-        let result =
-            self.conversation_client
-                .run_turn(&mut self.conversation_history, input, |event| {
-                    Self::handle_stream_event(
-                        renderer,
-                        event,
-                        &mut stream_spinner,
-                        &mut tool_spinner,
-                        &mut saw_text,
-                        &mut turn_usage,
-                        out,
-                    );
-                });
-
-        let summary = match result {
-            Ok(summary) => summary,
-            Err(error) => {
-                stream_spinner.fail(
-                    "Streaming response failed",
-                    self.renderer.color_theme(),
-                    out,
-                )?;
-                return Err(io::Error::other(error));
-            }
-        };
-        self.state.last_usage = summary.usage.clone();
-        if saw_text {
-            writeln!(out)?;
-        } else {
-            stream_spinner.finish("Streaming response", self.renderer.color_theme(), out)?;
-        }
-
-        self.write_turn_output(&summary, out)?;
-        let _ = turn_usage;
-        Ok(())
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use std::path::PathBuf;
-
-    use crate::args::{OutputFormat, PermissionMode};
-
-    use super::{CommandResult, SessionConfig, SlashCommand};
-
-    #[test]
-    fn parses_required_slash_commands() {
-        assert_eq!(SlashCommand::parse("/help"), Some(SlashCommand::Help));
-        assert_eq!(SlashCommand::parse(" /status "), Some(SlashCommand::Status));
-        assert_eq!(
-            SlashCommand::parse("/compact now"),
-            Some(SlashCommand::Compact)
-        );
-    }
-
-    #[test]
-    fn help_output_lists_commands() {
-        let mut out = Vec::new();
-        let result = super::CliApp::handle_help(&mut out).expect("help succeeds");
-        assert_eq!(result, CommandResult::Continue);
-        let output = String::from_utf8_lossy(&out);
-        assert!(output.contains("/help"));
-        assert!(output.contains("/status"));
-        assert!(output.contains("/compact"));
-    }
-
-    #[test]
-    fn session_state_tracks_config_values() {
-        let config = SessionConfig {
-            model: "claude".into(),
-            permission_mode: PermissionMode::WorkspaceWrite,
-            config: Some(PathBuf::from("settings.toml")),
-            output_format: OutputFormat::Text,
-        };
-
-        assert_eq!(config.model, "claude");
-        assert_eq!(config.permission_mode, PermissionMode::WorkspaceWrite);
-        assert_eq!(config.config, Some(PathBuf::from("settings.toml")));
-    }
-}
@@ -1,102 +0,0 @@
-use std::path::PathBuf;
-
-use clap::{Parser, Subcommand, ValueEnum};
-
-#[derive(Debug, Clone, Parser, PartialEq, Eq)]
-#[command(
-    name = "rusty-claude-cli",
-    version,
-    about = "Rust Claude CLI prototype"
-)]
-pub struct Cli {
-    #[arg(long, default_value = "claude-opus-4-6")]
-    pub model: String,
-
-    #[arg(long, value_enum, default_value_t = PermissionMode::WorkspaceWrite)]
-    pub permission_mode: PermissionMode,
-
-    #[arg(long)]
-    pub config: Option<PathBuf>,
-
-    #[arg(long, value_enum, default_value_t = OutputFormat::Text)]
-    pub output_format: OutputFormat,
-
-    #[command(subcommand)]
-    pub command: Option<Command>,
-}
-
-#[derive(Debug, Clone, Subcommand, PartialEq, Eq)]
-pub enum Command {
-    /// Read upstream TS sources and print extracted counts
-    DumpManifests,
-    /// Print the current bootstrap phase skeleton
-    BootstrapPlan,
-    /// Start the OAuth login flow
-    Login,
-    /// Clear saved OAuth credentials
-    Logout,
-    /// Run a non-interactive prompt and exit
-    Prompt { prompt: Vec<String> },
-}
-
-#[derive(Debug, Clone, Copy, ValueEnum, PartialEq, Eq)]
-pub enum PermissionMode {
-    ReadOnly,
-    WorkspaceWrite,
-    DangerFullAccess,
-}
-
-#[derive(Debug, Clone, Copy, ValueEnum, PartialEq, Eq)]
-pub enum OutputFormat {
-    Text,
-    Json,
-    Ndjson,
-}
-
-#[cfg(test)]
-mod tests {
-    use clap::Parser;
-
-    use super::{Cli, Command, OutputFormat, PermissionMode};
-
-    #[test]
-    fn parses_requested_flags() {
-        let cli = Cli::parse_from([
-            "rusty-claude-cli",
-            "--model",
-            "claude-3-5-haiku",
-            "--permission-mode",
-            "read-only",
-            "--config",
-            "/tmp/config.toml",
-            "--output-format",
-            "ndjson",
-            "prompt",
-            "hello",
-            "world",
-        ]);
-
-        assert_eq!(cli.model, "claude-3-5-haiku");
-        assert_eq!(cli.permission_mode, PermissionMode::ReadOnly);
-        assert_eq!(
-            cli.config.as_deref(),
-            Some(std::path::Path::new("/tmp/config.toml"))
-        );
-        assert_eq!(cli.output_format, OutputFormat::Ndjson);
-        assert_eq!(
-            cli.command,
-            Some(Command::Prompt {
-                prompt: vec!["hello".into(), "world".into()]
-            })
-        );
-    }
-
-    #[test]
-    fn parses_login_and_logout_commands() {
-        let login = Cli::parse_from(["rusty-claude-cli", "login"]);
-        assert_eq!(login.command, Some(Command::Login));
-
-        let logout = Cli::parse_from(["rusty-claude-cli", "logout"]);
-        assert_eq!(logout.command, Some(Command::Logout));
-    }
-}
@@ -1,15 +1,15 @@
 use std::fs;
 use std::path::{Path, PathBuf};

-const STARTER_CLAUDE_JSON: &str = concat!(
+const STARTER_CLAW_JSON: &str = concat!(
    "{\n",
    "  \"permissions\": {\n",
-    "    \"defaultMode\": \"acceptEdits\"\n",
+    "    \"defaultMode\": \"dontAsk\"\n",
    "  }\n",
    "}\n",
 );
-const GITIGNORE_COMMENT: &str = "# Claude Code local artifacts";
-const GITIGNORE_ENTRIES: [&str; 2] = [".claude/settings.local.json", ".claude/sessions/"];
+const GITIGNORE_COMMENT: &str = "# Claw Code local artifacts";
+const GITIGNORE_ENTRIES: [&str; 2] = [".claw/settings.local.json", ".claw/sessions/"];

 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub(crate) enum InitStatus {
@@ -80,16 +80,16 @@ struct RepoDetection {
 pub(crate) fn initialize_repo(cwd: &Path) -> Result<InitReport, Box<dyn std::error::Error>> {
    let mut artifacts = Vec::new();

-    let claude_dir = cwd.join(".claude");
+    let claw_dir = cwd.join(".claw");
    artifacts.push(InitArtifact {
-        name: ".claude/",
-        status: ensure_dir(&claude_dir)?,
+        name: ".claw/",
+        status: ensure_dir(&claw_dir)?,
    });

-    let claude_json = cwd.join(".claude.json");
+    let claw_json = cwd.join(".claw.json");
    artifacts.push(InitArtifact {
-        name: ".claude.json",
-        status: write_file_if_missing(&claude_json, STARTER_CLAUDE_JSON)?,
+        name: ".claw.json",
+        status: write_file_if_missing(&claw_json, STARTER_CLAW_JSON)?,
    });

    let gitignore = cwd.join(".gitignore");
@@ -164,7 +164,7 @@ pub(crate) fn render_init_claude_md(cwd: &Path) -> String {
    let mut lines = vec![
        "# CLAUDE.md".to_string(),
        String::new(),
-        "This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.".to_string(),
+        "This file provides guidance to Claw Code (clawcode.dev) when working with code in this repository.".to_string(),
        String::new(),
    ];

@@ -209,7 +209,7 @@ pub(crate) fn render_init_claude_md(cwd: &Path) -> String {

    lines.push("## Working agreement".to_string());
    lines.push("- Prefer small, reviewable changes and keep generated bootstrap files aligned with actual repo workflows.".to_string());
-    lines.push("- Keep shared defaults in `.claude.json`; reserve `.claude/settings.local.json` for machine-local overrides.".to_string());
+    lines.push("- Keep shared defaults in `.claw.json`; reserve `.claw/settings.local.json` for machine-local overrides.".to_string());
    lines.push("- Do not overwrite existing `CLAUDE.md` content automatically; update it intentionally when repo workflows change.".to_string());
    lines.push(String::new());

@@ -354,26 +354,27 @@ mod tests {

        let report = initialize_repo(&root).expect("init should succeed");
        let rendered = report.render();
-        assert!(rendered.contains(".claude/         created"));
-        assert!(rendered.contains(".claude.json     created"));
+        assert!(rendered.contains(".claw/"));
+        assert!(rendered.contains(".claw.json"));
+        assert!(rendered.contains("created"));
        assert!(rendered.contains(".gitignore       created"));
        assert!(rendered.contains("CLAUDE.md        created"));
-        assert!(root.join(".claude").is_dir());
-        assert!(root.join(".claude.json").is_file());
+        assert!(root.join(".claw").is_dir());
+        assert!(root.join(".claw.json").is_file());
        assert!(root.join("CLAUDE.md").is_file());
        assert_eq!(
-            fs::read_to_string(root.join(".claude.json")).expect("read claude json"),
+            fs::read_to_string(root.join(".claw.json")).expect("read claw json"),
            concat!(
                "{\n",
                "  \"permissions\": {\n",
-                "    \"defaultMode\": \"acceptEdits\"\n",
+                "    \"defaultMode\": \"dontAsk\"\n",
                "  }\n",
                "}\n",
            )
        );
        let gitignore = fs::read_to_string(root.join(".gitignore")).expect("read gitignore");
-        assert!(gitignore.contains(".claude/settings.local.json"));
-        assert!(gitignore.contains(".claude/sessions/"));
+        assert!(gitignore.contains(".claw/settings.local.json"));
+        assert!(gitignore.contains(".claw/sessions/"));
        let claude_md = fs::read_to_string(root.join("CLAUDE.md")).expect("read claude md");
        assert!(claude_md.contains("Languages: Rust."));
        assert!(claude_md.contains("cargo clippy --workspace --all-targets -- -D warnings"));
@@ -386,8 +387,7 @@ mod tests {
        let root = temp_dir();
        fs::create_dir_all(&root).expect("create root");
        fs::write(root.join("CLAUDE.md"), "custom guidance\n").expect("write existing claude md");
-        fs::write(root.join(".gitignore"), ".claude/settings.local.json\n")
-            .expect("write gitignore");
+        fs::write(root.join(".gitignore"), ".claw/settings.local.json\n").expect("write gitignore");

        let first = initialize_repo(&root).expect("first init should succeed");
        assert!(first
@@ -395,8 +395,9 @@ mod tests {
            .contains("CLAUDE.md        skipped (already exists)"));
        let second = initialize_repo(&root).expect("second init should succeed");
        let second_rendered = second.render();
-        assert!(second_rendered.contains(".claude/         skipped (already exists)"));
-        assert!(second_rendered.contains(".claude.json     skipped (already exists)"));
+        assert!(second_rendered.contains(".claw/"));
+        assert!(second_rendered.contains(".claw.json"));
+        assert!(second_rendered.contains("skipped (already exists)"));
        assert!(second_rendered.contains(".gitignore       skipped (already exists)"));
        assert!(second_rendered.contains("CLAUDE.md        skipped (already exists)"));
        assert_eq!(
@@ -404,8 +405,8 @@ mod tests {
            "custom guidance\n"
        );
        let gitignore = fs::read_to_string(root.join(".gitignore")).expect("read gitignore");
-        assert_eq!(gitignore.matches(".claude/settings.local.json").count(), 1);
-        assert_eq!(gitignore.matches(".claude/sessions/").count(), 1);
+        assert_eq!(gitignore.matches(".claw/settings.local.json").count(), 1);
+        assert_eq!(gitignore.matches(".claw/sessions/").count(), 1);

        fs::remove_dir_all(root).expect("cleanup temp dir");
    }
@@ -1,166 +1,17 @@
+use std::borrow::Cow;
+use std::cell::RefCell;
+use std::collections::BTreeSet;
 use std::io::{self, IsTerminal, Write};

-use crossterm::cursor::{MoveDown, MoveToColumn, MoveUp};
-use crossterm::event::{self, Event, KeyCode, KeyEvent, KeyModifiers};
-use crossterm::queue;
-use crossterm::terminal::{disable_raw_mode, enable_raw_mode, Clear, ClearType};
-
-#[derive(Debug, Clone, PartialEq, Eq)]
-pub struct InputBuffer {
-    buffer: String,
-    cursor: usize,
-}
-
-impl InputBuffer {
-    #[must_use]
-    pub fn new() -> Self {
-        Self {
-            buffer: String::new(),
-            cursor: 0,
-        }
-    }
-
-    pub fn insert(&mut self, ch: char) {
-        self.buffer.insert(self.cursor, ch);
-        self.cursor += ch.len_utf8();
-    }
-
-    pub fn insert_newline(&mut self) {
-        self.insert('\n');
-    }
-
-    pub fn backspace(&mut self) {
-        if self.cursor == 0 {
-            return;
-        }
-
-        let previous = self.buffer[..self.cursor]
-            .char_indices()
-            .last()
-            .map_or(0, |(idx, _)| idx);
-        self.buffer.drain(previous..self.cursor);
-        self.cursor = previous;
-    }
-
-    pub fn move_left(&mut self) {
-        if self.cursor == 0 {
-            return;
-        }
-        self.cursor = self.buffer[..self.cursor]
-            .char_indices()
-            .last()
-            .map_or(0, |(idx, _)| idx);
-    }
-
-    pub fn move_right(&mut self) {
-        if self.cursor >= self.buffer.len() {
-            return;
-        }
-        if let Some(next) = self.buffer[self.cursor..].chars().next() {
-            self.cursor += next.len_utf8();
-        }
-    }
-
-    pub fn move_home(&mut self) {
-        self.cursor = 0;
-    }
-
-    pub fn move_end(&mut self) {
-        self.cursor = self.buffer.len();
-    }
-
-    #[must_use]
-    pub fn as_str(&self) -> &str {
-        &self.buffer
-    }
-
-    #[cfg(test)]
-    #[must_use]
-    pub fn cursor(&self) -> usize {
-        self.cursor
-    }
-
-    pub fn clear(&mut self) {
-        self.buffer.clear();
-        self.cursor = 0;
-    }
-
-    pub fn replace(&mut self, value: impl Into<String>) {
-        self.buffer = value.into();
-        self.cursor = self.buffer.len();
-    }
-
-    #[must_use]
-    fn current_command_prefix(&self) -> Option<&str> {
-        if self.cursor != self.buffer.len() {
-            return None;
-        }
-        let prefix = &self.buffer[..self.cursor];
-        if prefix.contains(char::is_whitespace) || !prefix.starts_with('/') {
-            return None;
-        }
-        Some(prefix)
-    }
-
-    pub fn complete_slash_command(&mut self, candidates: &[String]) -> bool {
-        let Some(prefix) = self.current_command_prefix() else {
-            return false;
-        };
-
-        let matches = candidates
-            .iter()
-            .filter(|candidate| candidate.starts_with(prefix))
-            .map(String::as_str)
-            .collect::<Vec<_>>();
-        if matches.is_empty() {
-            return false;
-        }
-
-        let replacement = longest_common_prefix(&matches);
-        if replacement == prefix {
-            return false;
-        }
-
-        self.replace(replacement);
-        true
-    }
-}
-
-#[derive(Debug, Clone, PartialEq, Eq)]
-pub struct RenderedBuffer {
-    lines: Vec<String>,
-    cursor_row: u16,
-    cursor_col: u16,
-}
-
-impl RenderedBuffer {
-    #[must_use]
-    pub fn line_count(&self) -> usize {
-        self.lines.len()
-    }
-
-    fn write(&self, out: &mut impl Write) -> io::Result<()> {
-        for (index, line) in self.lines.iter().enumerate() {
-            if index > 0 {
-                writeln!(out)?;
-            }
-            write!(out, "{line}")?;
-        }
-        Ok(())
-    }
-
-    #[cfg(test)]
-    #[must_use]
-    pub fn lines(&self) -> &[String] {
-        &self.lines
-    }
-
-    #[cfg(test)]
-    #[must_use]
-    pub fn cursor_position(&self) -> (u16, u16) {
-        (self.cursor_row, self.cursor_col)
-    }
-}
+use rustyline::completion::{Completer, Pair};
+use rustyline::error::ReadlineError;
+use rustyline::highlight::{CmdKind, Highlighter};
+use rustyline::hint::Hinter;
+use rustyline::history::DefaultHistory;
+use rustyline::validate::Validator;
+use rustyline::{
+    Cmd, CompletionType, Config, Context, EditMode, Editor, Helper, KeyCode, KeyEvent, Modifiers,
+};

 #[derive(Debug, Clone, PartialEq, Eq)]
 pub enum ReadOutcome {
@@ -169,25 +20,105 @@ pub enum ReadOutcome {
    Exit,
 }

+struct SlashCommandHelper {
+    completions: Vec<String>,
+    current_line: RefCell<String>,
+}
+
+impl SlashCommandHelper {
+    fn new(completions: Vec<String>) -> Self {
+        Self {
+            completions: normalize_completions(completions),
+            current_line: RefCell::new(String::new()),
+        }
+    }
+
+    fn reset_current_line(&self) {
+        self.current_line.borrow_mut().clear();
+    }
+
+    fn current_line(&self) -> String {
+        self.current_line.borrow().clone()
+    }
+
+    fn set_current_line(&self, line: &str) {
+        let mut current = self.current_line.borrow_mut();
+        current.clear();
+        current.push_str(line);
+    }
+
+    fn set_completions(&mut self, completions: Vec<String>) {
+        self.completions = normalize_completions(completions);
+    }
+}
+
+impl Completer for SlashCommandHelper {
+    type Candidate = Pair;
+
+    fn complete(
+        &self,
+        line: &str,
+        pos: usize,
+        _ctx: &Context<'_>,
+    ) -> rustyline::Result<(usize, Vec<Self::Candidate>)> {
+        let Some(prefix) = slash_command_prefix(line, pos) else {
+            return Ok((0, Vec::new()));
+        };
+
+        let matches = self
+            .completions
+            .iter()
+            .filter(|candidate| candidate.starts_with(prefix))
+            .map(|candidate| Pair {
+                display: candidate.clone(),
+                replacement: candidate.clone(),
+            })
+            .collect();
+
+        Ok((0, matches))
+    }
+}
+
+impl Hinter for SlashCommandHelper {
+    type Hint = String;
+}
+
+impl Highlighter for SlashCommandHelper {
+    fn highlight<'l>(&self, line: &'l str, _pos: usize) -> Cow<'l, str> {
+        self.set_current_line(line);
+        Cow::Borrowed(line)
+    }
+
+    fn highlight_char(&self, line: &str, _pos: usize, _kind: CmdKind) -> bool {
+        self.set_current_line(line);
+        false
+    }
+}
+
+impl Validator for SlashCommandHelper {}
+impl Helper for SlashCommandHelper {}
+
 pub struct LineEditor {
    prompt: String,
-    continuation_prompt: String,
-    history: Vec<String>,
-    history_index: Option<usize>,
-    draft: Option<String>,
-    completions: Vec<String>,
+    editor: Editor<SlashCommandHelper, DefaultHistory>,
 }

 impl LineEditor {
    #[must_use]
    pub fn new(prompt: impl Into<String>, completions: Vec<String>) -> Self {
+        let config = Config::builder()
+            .completion_type(CompletionType::List)
+            .edit_mode(EditMode::Emacs)
+            .build();
+        let mut editor = Editor::<SlashCommandHelper, DefaultHistory>::with_config(config)
+            .expect("rustyline editor should initialize");
+        editor.set_helper(Some(SlashCommandHelper::new(completions)));
+        editor.bind_sequence(KeyEvent(KeyCode::Char('J'), Modifiers::CTRL), Cmd::Newline);
+        editor.bind_sequence(KeyEvent(KeyCode::Enter, Modifiers::SHIFT), Cmd::Newline);
+
        Self {
            prompt: prompt.into(),
-            continuation_prompt: String::from("> "),
-            history: Vec::new(),
-            history_index: None,
-            draft: None,
-            completions,
+            editor,
        }
    }

@@ -196,9 +127,14 @@ impl LineEditor {
        if entry.trim().is_empty() {
            return;
        }
-        self.history.push(entry);
-        self.history_index = None;
-        self.draft = None;
+
+        let _ = self.editor.add_history_entry(entry);
+    }
+
+    pub fn set_completions(&mut self, completions: Vec<String>) {
+        if let Some(helper) = self.editor.helper_mut() {
+            helper.set_completions(completions);
+        }
    }

    pub fn read_line(&mut self) -> io::Result<ReadOutcome> {
@@ -206,45 +142,43 @@ impl LineEditor {
            return self.read_line_fallback();
        }

-        enable_raw_mode()?;
-        let mut stdout = io::stdout();
-        let mut input = InputBuffer::new();
-        let mut rendered_lines = 1usize;
-        self.redraw(&mut stdout, &input, rendered_lines)?;
+        if let Some(helper) = self.editor.helper_mut() {
+            helper.reset_current_line();
+        }

-        loop {
-            let event = event::read()?;
-            if let Event::Key(key) = event {
-                match self.handle_key(key, &mut input) {
-                    EditorAction::Continue => {
-                        rendered_lines = self.redraw(&mut stdout, &input, rendered_lines)?;
-                    }
-                    EditorAction::Submit => {
-                        disable_raw_mode()?;
-                        writeln!(stdout)?;
-                        self.history_index = None;
-                        self.draft = None;
-                        return Ok(ReadOutcome::Submit(input.as_str().to_owned()));
-                    }
-                    EditorAction::Cancel => {
-                        disable_raw_mode()?;
-                        writeln!(stdout)?;
-                        self.history_index = None;
-                        self.draft = None;
-                        return Ok(ReadOutcome::Cancel);
-                    }
-                    EditorAction::Exit => {
-                        disable_raw_mode()?;
-                        writeln!(stdout)?;
-                        self.history_index = None;
-                        self.draft = None;
-                        return Ok(ReadOutcome::Exit);
-                    }
+        match self.editor.readline(&self.prompt) {
+            Ok(line) => Ok(ReadOutcome::Submit(line)),
+            Err(ReadlineError::Interrupted) => {
+                let has_input = !self.current_line().is_empty();
+                self.finish_interrupted_read()?;
+                if has_input {
+                    Ok(ReadOutcome::Cancel)
+                } else {
+                    Ok(ReadOutcome::Exit)
                }
            }
+            Err(ReadlineError::Eof) => {
+                self.finish_interrupted_read()?;
+                Ok(ReadOutcome::Exit)
+            }
+            Err(error) => Err(io::Error::other(error)),
        }
    }

+    fn current_line(&self) -> String {
+        self.editor
+            .helper()
+            .map_or_else(String::new, SlashCommandHelper::current_line)
+    }
+
+    fn finish_interrupted_read(&mut self) -> io::Result<()> {
+        if let Some(helper) = self.editor.helper_mut() {
+            helper.reset_current_line();
+        }
+        let mut stdout = io::stdout();
+        writeln!(stdout)
+    }
+
    fn read_line_fallback(&self) -> io::Result<ReadOutcome> {
        let mut stdout = io::stdout();
        write!(stdout, "{}", self.prompt)?;
@@ -261,388 +195,136 @@ impl LineEditor {
        }
        Ok(ReadOutcome::Submit(buffer))
    }
-
-    #[allow(clippy::too_many_lines)]
-    fn handle_key(&mut self, key: KeyEvent, input: &mut InputBuffer) -> EditorAction {
-        match key {
-            KeyEvent {
-                code: KeyCode::Char('c'),
-                modifiers,
-                ..
-            } if modifiers.contains(KeyModifiers::CONTROL) => {
-                if input.as_str().is_empty() {
-                    EditorAction::Exit
-                } else {
-                    input.clear();
-                    self.history_index = None;
-                    self.draft = None;
-                    EditorAction::Cancel
-                }
-            }
-            KeyEvent {
-                code: KeyCode::Char('j'),
-                modifiers,
-                ..
-            } if modifiers.contains(KeyModifiers::CONTROL) => {
-                input.insert_newline();
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Enter,
-                modifiers,
-                ..
-            } if modifiers.contains(KeyModifiers::SHIFT) => {
-                input.insert_newline();
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Enter,
-                ..
-            } => EditorAction::Submit,
-            KeyEvent {
-                code: KeyCode::Backspace,
-                ..
-            } => {
-                input.backspace();
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Left,
-                ..
-            } => {
-                input.move_left();
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Right,
-                ..
-            } => {
-                input.move_right();
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Up, ..
-            } => {
-                self.navigate_history_up(input);
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Down,
-                ..
-            } => {
-                self.navigate_history_down(input);
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Tab, ..
-            } => {
-                input.complete_slash_command(&self.completions);
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Home,
-                ..
-            } => {
-                input.move_home();
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::End, ..
-            } => {
-                input.move_end();
-                EditorAction::Continue
-            }
-            KeyEvent {
-                code: KeyCode::Esc, ..
-            } => {
-                input.clear();
-                self.history_index = None;
-                self.draft = None;
-                EditorAction::Cancel
-            }
-            KeyEvent {
-                code: KeyCode::Char(ch),
-                modifiers,
-                ..
-            } if modifiers.is_empty() || modifiers == KeyModifiers::SHIFT => {
-                input.insert(ch);
-                self.history_index = None;
-                self.draft = None;
-                EditorAction::Continue
-            }
-            _ => EditorAction::Continue,
-        }
-    }
-
-    fn navigate_history_up(&mut self, input: &mut InputBuffer) {
-        if self.history.is_empty() {
-            return;
-        }
-
-        match self.history_index {
-            Some(0) => {}
-            Some(index) => {
-                let next_index = index - 1;
-                input.replace(self.history[next_index].clone());
-                self.history_index = Some(next_index);
-            }
-            None => {
-                self.draft = Some(input.as_str().to_owned());
-                let next_index = self.history.len() - 1;
-                input.replace(self.history[next_index].clone());
-                self.history_index = Some(next_index);
-            }
-        }
-    }
-
-    fn navigate_history_down(&mut self, input: &mut InputBuffer) {
-        let Some(index) = self.history_index else {
-            return;
-        };
-
-        if index + 1 < self.history.len() {
-            let next_index = index + 1;
-            input.replace(self.history[next_index].clone());
-            self.history_index = Some(next_index);
-            return;
-        }
-
-        input.replace(self.draft.take().unwrap_or_default());
-        self.history_index = None;
-    }
-
-    fn redraw(
-        &self,
-        out: &mut impl Write,
-        input: &InputBuffer,
-        previous_line_count: usize,
-    ) -> io::Result<usize> {
-        let rendered = render_buffer(&self.prompt, &self.continuation_prompt, input);
-        if previous_line_count > 1 {
-            queue!(out, MoveUp(saturating_u16(previous_line_count - 1)))?;
-        }
-        queue!(out, MoveToColumn(0), Clear(ClearType::FromCursorDown),)?;
-        rendered.write(out)?;
-        queue!(
-            out,
-            MoveUp(saturating_u16(rendered.line_count().saturating_sub(1))),
-            MoveToColumn(0),
-        )?;
-        if rendered.cursor_row > 0 {
-            queue!(out, MoveDown(rendered.cursor_row))?;
-        }
-        queue!(out, MoveToColumn(rendered.cursor_col))?;
-        out.flush()?;
-        Ok(rendered.line_count())
-    }
 }

-#[derive(Debug, Clone, Copy, PartialEq, Eq)]
-enum EditorAction {
-    Continue,
-    Submit,
-    Cancel,
-    Exit,
+fn slash_command_prefix(line: &str, pos: usize) -> Option<&str> {
+    if pos != line.len() {
+        return None;
+    }
+
+    let prefix = &line[..pos];
+    if !prefix.starts_with('/') {
+        return None;
+    }
+
+    Some(prefix)
 }

-#[must_use]
-pub fn render_buffer(
-    prompt: &str,
-    continuation_prompt: &str,
-    input: &InputBuffer,
-) -> RenderedBuffer {
-    let before_cursor = &input.as_str()[..input.cursor];
-    let cursor_row = saturating_u16(before_cursor.chars().filter(|ch| *ch == '\n').count());
-    let cursor_line = before_cursor.rsplit('\n').next().unwrap_or_default();
-    let cursor_prompt = if cursor_row == 0 {
-        prompt
-    } else {
-        continuation_prompt
-    };
-    let cursor_col = saturating_u16(cursor_prompt.chars().count() + cursor_line.chars().count());
-
-    let mut lines = Vec::new();
-    for (index, line) in input.as_str().split('\n').enumerate() {
-        let prefix = if index == 0 {
-            prompt
-        } else {
-            continuation_prompt
-        };
-        lines.push(format!("{prefix}{line}"));
-    }
-    if lines.is_empty() {
-        lines.push(prompt.to_string());
-    }
-
-    RenderedBuffer {
-        lines,
-        cursor_row,
-        cursor_col,
-    }
-}
-
-#[must_use]
-fn longest_common_prefix(values: &[&str]) -> String {
-    let Some(first) = values.first() else {
-        return String::new();
-    };
-
-    let mut prefix = (*first).to_string();
-    for value in values.iter().skip(1) {
-        while !value.starts_with(&prefix) {
-            prefix.pop();
-            if prefix.is_empty() {
-                break;
-            }
-        }
-    }
-    prefix
-}
-
-#[must_use]
-fn saturating_u16(value: usize) -> u16 {
-    u16::try_from(value).unwrap_or(u16::MAX)
+fn normalize_completions(completions: Vec<String>) -> Vec<String> {
+    let mut seen = BTreeSet::new();
+    completions
+        .into_iter()
+        .filter(|candidate| candidate.starts_with('/'))
+        .filter(|candidate| seen.insert(candidate.clone()))
+        .collect()
 }

 #[cfg(test)]
 mod tests {
-    use super::{render_buffer, InputBuffer, LineEditor};
-    use crossterm::event::{KeyCode, KeyEvent, KeyModifiers};
+    use super::{slash_command_prefix, LineEditor, SlashCommandHelper};
+    use rustyline::completion::Completer;
+    use rustyline::highlight::Highlighter;
+    use rustyline::history::{DefaultHistory, History};
+    use rustyline::Context;

-    fn key(code: KeyCode) -> KeyEvent {
-        KeyEvent::new(code, KeyModifiers::NONE)
+    #[test]
+    fn extracts_terminal_slash_command_prefixes_with_arguments() {
+        assert_eq!(slash_command_prefix("/he", 3), Some("/he"));
+        assert_eq!(slash_command_prefix("/help me", 8), Some("/help me"));
+        assert_eq!(
+            slash_command_prefix("/session switch ses", 19),
+            Some("/session switch ses")
+        );
+        assert_eq!(slash_command_prefix("hello", 5), None);
+        assert_eq!(slash_command_prefix("/help", 2), None);
    }

    #[test]
-    fn supports_basic_line_editing() {
-        let mut input = InputBuffer::new();
-        input.insert('h');
-        input.insert('i');
-        input.move_end();
-        input.insert_newline();
-        input.insert('x');
-
-        assert_eq!(input.as_str(), "hi\nx");
-        assert_eq!(input.cursor(), 4);
-
-        input.move_left();
-        input.backspace();
-        assert_eq!(input.as_str(), "hix");
-        assert_eq!(input.cursor(), 2);
-    }
-
-    #[test]
-    fn completes_unique_slash_command() {
-        let mut input = InputBuffer::new();
-        for ch in "/he".chars() {
-            input.insert(ch);
-        }
-
-        assert!(input.complete_slash_command(&[
+    fn completes_matching_slash_commands() {
+        let helper = SlashCommandHelper::new(vec![
            "/help".to_string(),
            "/hello".to_string(),
            "/status".to_string(),
-        ]));
-        assert_eq!(input.as_str(), "/hel");
+        ]);
+        let history = DefaultHistory::new();
+        let ctx = Context::new(&history);
+        let (start, matches) = helper
+            .complete("/he", 3, &ctx)
+            .expect("completion should work");

-        assert!(input.complete_slash_command(&["/help".to_string(), "/status".to_string()]));
-        assert_eq!(input.as_str(), "/help");
-    }
-
-    #[test]
-    fn ignores_completion_when_prefix_is_not_a_slash_command() {
-        let mut input = InputBuffer::new();
-        for ch in "hello".chars() {
-            input.insert(ch);
-        }
-
-        assert!(!input.complete_slash_command(&["/help".to_string()]));
-        assert_eq!(input.as_str(), "hello");
-    }
-
-    #[test]
-    fn history_navigation_restores_current_draft() {
-        let mut editor = LineEditor::new("› ", vec![]);
-        editor.push_history("/help");
-        editor.push_history("status report");
-
-        let mut input = InputBuffer::new();
-        for ch in "draft".chars() {
-            input.insert(ch);
-        }
-
-        let _ = editor.handle_key(key(KeyCode::Up), &mut input);
-        assert_eq!(input.as_str(), "status report");
-
-        let _ = editor.handle_key(key(KeyCode::Up), &mut input);
-        assert_eq!(input.as_str(), "/help");
-
-        let _ = editor.handle_key(key(KeyCode::Down), &mut input);
-        assert_eq!(input.as_str(), "status report");
-
-        let _ = editor.handle_key(key(KeyCode::Down), &mut input);
-        assert_eq!(input.as_str(), "draft");
-    }
-
-    #[test]
-    fn tab_key_completes_from_editor_candidates() {
-        let mut editor = LineEditor::new(
-            "› ",
-            vec![
-                "/help".to_string(),
-                "/status".to_string(),
-                "/session".to_string(),
-            ],
-        );
-        let mut input = InputBuffer::new();
-        for ch in "/st".chars() {
-            input.insert(ch);
-        }
-
-        let _ = editor.handle_key(key(KeyCode::Tab), &mut input);
-        assert_eq!(input.as_str(), "/status");
-    }
-
-    #[test]
-    fn renders_multiline_buffers_with_continuation_prompt() {
-        let mut input = InputBuffer::new();
-        for ch in "hello\nworld".chars() {
-            if ch == '\n' {
-                input.insert_newline();
-            } else {
-                input.insert(ch);
-            }
-        }
-
-        let rendered = render_buffer("› ", "> ", &input);
+        assert_eq!(start, 0);
        assert_eq!(
-            rendered.lines(),
-            &["› hello".to_string(), "> world".to_string()]
+            matches
+                .into_iter()
+                .map(|candidate| candidate.replacement)
+                .collect::<Vec<_>>(),
+            vec!["/help".to_string(), "/hello".to_string()]
        );
-        assert_eq!(rendered.cursor_position(), (1, 7));
    }

    #[test]
-    fn ctrl_c_exits_only_when_buffer_is_empty() {
-        let mut editor = LineEditor::new("› ", vec![]);
-        let mut empty = InputBuffer::new();
-        assert!(matches!(
-            editor.handle_key(
-                KeyEvent::new(KeyCode::Char('c'), KeyModifiers::CONTROL),
-                &mut empty,
-            ),
-            super::EditorAction::Exit
-        ));
+    fn completes_matching_slash_command_arguments() {
+        let helper = SlashCommandHelper::new(vec![
+            "/model".to_string(),
+            "/model opus".to_string(),
+            "/model sonnet".to_string(),
+            "/session switch alpha".to_string(),
+        ]);
+        let history = DefaultHistory::new();
+        let ctx = Context::new(&history);
+        let (start, matches) = helper
+            .complete("/model o", 8, &ctx)
+            .expect("completion should work");

-        let mut filled = InputBuffer::new();
-        filled.insert('x');
-        assert!(matches!(
-            editor.handle_key(
-                KeyEvent::new(KeyCode::Char('c'), KeyModifiers::CONTROL),
-                &mut filled,
-            ),
-            super::EditorAction::Cancel
-        ));
-        assert!(filled.as_str().is_empty());
+        assert_eq!(start, 0);
+        assert_eq!(
+            matches
+                .into_iter()
+                .map(|candidate| candidate.replacement)
+                .collect::<Vec<_>>(),
+            vec!["/model opus".to_string()]
+        );
+    }
+
+    #[test]
+    fn ignores_non_slash_command_completion_requests() {
+        let helper = SlashCommandHelper::new(vec!["/help".to_string()]);
+        let history = DefaultHistory::new();
+        let ctx = Context::new(&history);
+        let (_, matches) = helper
+            .complete("hello", 5, &ctx)
+            .expect("completion should work");
+
+        assert!(matches.is_empty());
+    }
+
+    #[test]
+    fn tracks_current_buffer_through_highlighter() {
+        let helper = SlashCommandHelper::new(Vec::new());
+        let _ = helper.highlight("draft", 5);
+
+        assert_eq!(helper.current_line(), "draft");
+    }
+
+    #[test]
+    fn push_history_ignores_blank_entries() {
+        let mut editor = LineEditor::new("> ", vec!["/help".to_string()]);
+        editor.push_history("   ");
+        editor.push_history("/help");
+
+        assert_eq!(editor.editor.history().len(), 1);
+    }
+
+    #[test]
+    fn set_completions_replaces_and_normalizes_candidates() {
+        let mut editor = LineEditor::new("> ", vec!["/help".to_string()]);
+        editor.set_completions(vec![
+            "/model opus".to_string(),
+            "/model opus".to_string(),
+            "status".to_string(),
+        ]);
+
+        let helper = editor.editor.helper().expect("helper should exist");
+        assert_eq!(helper.completions, vec!["/model opus".to_string()]);
    }
 }
@@ -1,7 +1,5 @@
 use std::fmt::Write as FmtWrite;
 use std::io::{self, Write};
-use std::thread;
-use std::time::Duration;

 use crossterm::cursor::{MoveToColumn, RestorePosition, SavePosition};
 use crossterm::style::{Color, Print, ResetColor, SetForegroundColor, Stylize};
@@ -22,6 +20,7 @@ pub struct ColorTheme {
    link: Color,
    quote: Color,
    table_border: Color,
+    code_block_border: Color,
    spinner_active: Color,
    spinner_done: Color,
    spinner_failed: Color,
@@ -37,6 +36,7 @@ impl Default for ColorTheme {
            link: Color::Blue,
            quote: Color::DarkGrey,
            table_border: Color::DarkCyan,
+            code_block_border: Color::DarkGrey,
            spinner_active: Color::Blue,
            spinner_done: Color::Green,
            spinner_failed: Color::Red,
@@ -154,33 +154,64 @@ impl TableState {
 struct RenderState {
    emphasis: usize,
    strong: usize,
+    heading_level: Option<u8>,
    quote: usize,
    list_stack: Vec<ListKind>,
+    link_stack: Vec<LinkState>,
    table: Option<TableState>,
 }

+#[derive(Debug, Clone, PartialEq, Eq)]
+struct LinkState {
+    destination: String,
+    text: String,
+}
+
 impl RenderState {
    fn style_text(&self, text: &str, theme: &ColorTheme) -> String {
-        let mut styled = text.to_string();
-        if self.strong > 0 {
-            styled = format!("{}", styled.bold().with(theme.strong));
+        let mut style = text.stylize();
+
+        if matches!(self.heading_level, Some(1 | 2)) || self.strong > 0 {
+            style = style.bold();
        }
        if self.emphasis > 0 {
-            styled = format!("{}", styled.italic().with(theme.emphasis));
+            style = style.italic();
        }
+
+        if let Some(level) = self.heading_level {
+            style = match level {
+                1 => style.with(theme.heading),
+                2 => style.white(),
+                3 => style.with(Color::Blue),
+                _ => style.with(Color::Grey),
+            };
+        } else if self.strong > 0 {
+            style = style.with(theme.strong);
+        } else if self.emphasis > 0 {
+            style = style.with(theme.emphasis);
+        }
+
        if self.quote > 0 {
-            styled = format!("{}", styled.with(theme.quote));
+            style = style.with(theme.quote);
        }
-        styled
+
+        format!("{style}")
    }

-    fn capture_target_mut<'a>(&'a mut self, output: &'a mut String) -> &'a mut String {
-        if let Some(table) = self.table.as_mut() {
-            &mut table.current_cell
+    fn append_raw(&mut self, output: &mut String, text: &str) {
+        if let Some(link) = self.link_stack.last_mut() {
+            link.text.push_str(text);
+        } else if let Some(table) = self.table.as_mut() {
+            table.current_cell.push_str(text);
        } else {
-            output
+            output.push_str(text);
        }
    }
+
+    fn append_styled(&mut self, output: &mut String, text: &str, theme: &ColorTheme) {
+        let styled = self.style_text(text, theme);
+        self.append_raw(output, &styled);
+    }
 }

 #[derive(Debug)]
@@ -218,13 +249,14 @@ impl TerminalRenderer {

    #[must_use]
    pub fn render_markdown(&self, markdown: &str) -> String {
+        let normalized = normalize_nested_fences(markdown);
        let mut output = String::new();
        let mut state = RenderState::default();
        let mut code_language = String::new();
        let mut code_buffer = String::new();
        let mut in_code_block = false;

-        for event in Parser::new_ext(markdown, Options::all()) {
+        for event in Parser::new_ext(&normalized, Options::all()) {
            self.render_event(
                event,
                &mut state,
@@ -238,6 +270,11 @@ impl TerminalRenderer {
        output.trim_end().to_string()
    }

+    #[must_use]
+    pub fn markdown_to_ansi(&self, markdown: &str) -> String {
+        self.render_markdown(markdown)
+    }
+
    #[allow(clippy::too_many_lines)]
    fn render_event(
        &self,
@@ -249,15 +286,21 @@ impl TerminalRenderer {
        in_code_block: &mut bool,
    ) {
        match event {
-            Event::Start(Tag::Heading { level, .. }) => self.start_heading(level as u8, output),
-            Event::End(TagEnd::Heading(..) | TagEnd::Paragraph) => output.push_str("\n\n"),
+            Event::Start(Tag::Heading { level, .. }) => {
+                Self::start_heading(state, level as u8, output);
+            }
+            Event::End(TagEnd::Paragraph) => output.push_str("\n\n"),
            Event::Start(Tag::BlockQuote(..)) => self.start_quote(state, output),
            Event::End(TagEnd::BlockQuote(..)) => {
                state.quote = state.quote.saturating_sub(1);
                output.push('\n');
            }
+            Event::End(TagEnd::Heading(..)) => {
+                state.heading_level = None;
+                output.push_str("\n\n");
+            }
            Event::End(TagEnd::Item) | Event::SoftBreak | Event::HardBreak => {
-                state.capture_target_mut(output).push('\n');
+                state.append_raw(output, "\n");
            }
            Event::Start(Tag::List(first_item)) => {
                let kind = match first_item {
@@ -293,41 +336,52 @@ impl TerminalRenderer {
            Event::Code(code) => {
                let rendered =
                    format!("{}", format!("`{code}`").with(self.color_theme.inline_code));
-                state.capture_target_mut(output).push_str(&rendered);
+                state.append_raw(output, &rendered);
            }
            Event::Rule => output.push_str("---\n"),
            Event::Text(text) => {
                self.push_text(text.as_ref(), state, output, code_buffer, *in_code_block);
            }
            Event::Html(html) | Event::InlineHtml(html) => {
-                state.capture_target_mut(output).push_str(&html);
+                state.append_raw(output, &html);
            }
            Event::FootnoteReference(reference) => {
-                let _ = write!(state.capture_target_mut(output), "[{reference}]");
+                state.append_raw(output, &format!("[{reference}]"));
            }
            Event::TaskListMarker(done) => {
-                state
-                    .capture_target_mut(output)
-                    .push_str(if done { "[x] " } else { "[ ] " });
+                state.append_raw(output, if done { "[x] " } else { "[ ] " });
            }
            Event::InlineMath(math) | Event::DisplayMath(math) => {
-                state.capture_target_mut(output).push_str(&math);
+                state.append_raw(output, &math);
            }
            Event::Start(Tag::Link { dest_url, .. }) => {
-                let rendered = format!(
-                    "{}",
-                    format!("[{dest_url}]")
-                        .underlined()
-                        .with(self.color_theme.link)
-                );
-                state.capture_target_mut(output).push_str(&rendered);
+                state.link_stack.push(LinkState {
+                    destination: dest_url.to_string(),
+                    text: String::new(),
+                });
+            }
+            Event::End(TagEnd::Link) => {
+                if let Some(link) = state.link_stack.pop() {
+                    let label = if link.text.is_empty() {
+                        link.destination.clone()
+                    } else {
+                        link.text
+                    };
+                    let rendered = format!(
+                        "{}",
+                        format!("[{label}]({})", link.destination)
+                            .underlined()
+                            .with(self.color_theme.link)
+                    );
+                    state.append_raw(output, &rendered);
+                }
            }
            Event::Start(Tag::Image { dest_url, .. }) => {
                let rendered = format!(
                    "{}",
                    format!("[image:{dest_url}]").with(self.color_theme.link)
                );
-                state.capture_target_mut(output).push_str(&rendered);
+                state.append_raw(output, &rendered);
            }
            Event::Start(Tag::Table(..)) => state.table = Some(TableState::default()),
            Event::End(TagEnd::Table) => {
@@ -369,19 +423,15 @@ impl TerminalRenderer {
                }
            }
            Event::Start(Tag::Paragraph | Tag::MetadataBlock(..) | _)
-            | Event::End(TagEnd::Link | TagEnd::Image | TagEnd::MetadataBlock(..) | _) => {}
+            | Event::End(TagEnd::Image | TagEnd::MetadataBlock(..) | _) => {}
        }
    }

-    fn start_heading(&self, level: u8, output: &mut String) {
-        output.push('\n');
-        let prefix = match level {
-            1 => "# ",
-            2 => "## ",
-            3 => "### ",
-            _ => "#### ",
-        };
-        let _ = write!(output, "{}", prefix.bold().with(self.color_theme.heading));
+    fn start_heading(state: &mut RenderState, level: u8, output: &mut String) {
+        state.heading_level = Some(level);
+        if !output.is_empty() {
+            output.push('\n');
+        }
    }

    fn start_quote(&self, state: &mut RenderState, output: &mut String) {
@@ -405,20 +455,27 @@ impl TerminalRenderer {
    }

    fn start_code_block(&self, code_language: &str, output: &mut String) {
-        if !code_language.is_empty() {
-            let _ = writeln!(
-                output,
-                "{}",
-                format!("╭─ {code_language}").with(self.color_theme.heading)
-            );
-        }
+        let label = if code_language.is_empty() {
+            "code".to_string()
+        } else {
+            code_language.to_string()
+        };
+        let _ = writeln!(
+            output,
+            "{}",
+            format!("╭─ {label}")
+                .bold()
+                .with(self.color_theme.code_block_border)
+        );
    }

    fn finish_code_block(&self, code_buffer: &str, code_language: &str, output: &mut String) {
        output.push_str(&self.highlight_code(code_buffer, code_language));
-        if !code_language.is_empty() {
-            let _ = write!(output, "{}", "╰─".with(self.color_theme.heading));
-        }
+        let _ = write!(
+            output,
+            "{}",
+            "╰─".bold().with(self.color_theme.code_block_border)
+        );
        output.push_str("\n\n");
    }

@@ -433,8 +490,7 @@ impl TerminalRenderer {
        if in_code_block {
            code_buffer.push_str(text);
        } else {
-            let rendered = state.style_text(text, &self.color_theme);
-            state.capture_target_mut(output).push_str(&rendered);
+            state.append_styled(output, text, &self.color_theme);
        }
    }

@@ -521,9 +577,10 @@ impl TerminalRenderer {
        for line in LinesWithEndings::from(code) {
            match syntax_highlighter.highlight_line(line, &self.syntax_set) {
                Ok(ranges) => {
-                    colored_output.push_str(&as_24_bit_terminal_escaped(&ranges[..], false));
+                    let escaped = as_24_bit_terminal_escaped(&ranges[..], false);
+                    colored_output.push_str(&apply_code_block_background(&escaped));
                }
-                Err(_) => colored_output.push_str(line),
+                Err(_) => colored_output.push_str(&apply_code_block_background(line)),
            }
        }

@@ -531,16 +588,296 @@ impl TerminalRenderer {
    }

    pub fn stream_markdown(&self, markdown: &str, out: &mut impl Write) -> io::Result<()> {
-        let rendered_markdown = self.render_markdown(markdown);
-        for chunk in rendered_markdown.split_inclusive(char::is_whitespace) {
-            write!(out, "{chunk}")?;
-            out.flush()?;
-            thread::sleep(Duration::from_millis(8));
+        let rendered_markdown = self.markdown_to_ansi(markdown);
+        write!(out, "{rendered_markdown}")?;
+        if !rendered_markdown.ends_with('\n') {
+            writeln!(out)?;
        }
-        writeln!(out)
+        out.flush()
    }
 }

+#[derive(Debug, Default, Clone, PartialEq, Eq)]
+pub struct MarkdownStreamState {
+    pending: String,
+}
+
+impl MarkdownStreamState {
+    #[must_use]
+    pub fn push(&mut self, renderer: &TerminalRenderer, delta: &str) -> Option<String> {
+        self.pending.push_str(delta);
+        let split = find_stream_safe_boundary(&self.pending)?;
+        let ready = self.pending[..split].to_string();
+        self.pending.drain(..split);
+        Some(renderer.markdown_to_ansi(&ready))
+    }
+
+    #[must_use]
+    pub fn flush(&mut self, renderer: &TerminalRenderer) -> Option<String> {
+        if self.pending.trim().is_empty() {
+            self.pending.clear();
+            None
+        } else {
+            let pending = std::mem::take(&mut self.pending);
+            Some(renderer.markdown_to_ansi(&pending))
+        }
+    }
+}
+
+fn apply_code_block_background(line: &str) -> String {
+    let trimmed = line.trim_end_matches('\n');
+    let trailing_newline = if trimmed.len() == line.len() {
+        ""
+    } else {
+        "\n"
+    };
+    let with_background = trimmed.replace("\u{1b}[0m", "\u{1b}[0;48;5;236m");
+    format!("\u{1b}[48;5;236m{with_background}\u{1b}[0m{trailing_newline}")
+}
+
+/// Pre-process raw markdown so that fenced code blocks whose body contains
+/// fence markers of equal or greater length are wrapped with a longer fence.
+///
+/// LLMs frequently emit triple-backtick code blocks that contain triple-backtick
+/// examples.  CommonMark (and pulldown-cmark) treats the inner marker as the
+/// closing fence, breaking the render.  This function detects the situation and
+/// upgrades the outer fence to use enough backticks (or tildes) that the inner
+/// markers become ordinary content.
+fn normalize_nested_fences(markdown: &str) -> String {
+    // A fence line is either "labeled" (has an info string ⇒ always an opener)
+    // or "bare" (no info string ⇒ could be opener or closer).
+    #[derive(Debug, Clone)]
+    struct FenceLine {
+        char: char,
+        len: usize,
+        has_info: bool,
+        indent: usize,
+    }
+
+    fn parse_fence_line(line: &str) -> Option<FenceLine> {
+        let trimmed = line.trim_end_matches('\n').trim_end_matches('\r');
+        let indent = trimmed.chars().take_while(|c| *c == ' ').count();
+        if indent > 3 {
+            return None;
+        }
+        let rest = &trimmed[indent..];
+        let ch = rest.chars().next()?;
+        if ch != '`' && ch != '~' {
+            return None;
+        }
+        let len = rest.chars().take_while(|c| *c == ch).count();
+        if len < 3 {
+            return None;
+        }
+        let after = &rest[len..];
+        if ch == '`' && after.contains('`') {
+            return None;
+        }
+        let has_info = !after.trim().is_empty();
+        Some(FenceLine {
+            char: ch,
+            len,
+            has_info,
+            indent,
+        })
+    }
+
+    let lines: Vec<&str> = markdown.split_inclusive('\n').collect();
+    // Handle final line that may lack trailing newline.
+    // split_inclusive already keeps the original chunks, including a
+    // final chunk without '\n' if the input doesn't end with one.
+
+    // First pass: classify every line.
+    let fence_info: Vec<Option<FenceLine>> = lines.iter().map(|l| parse_fence_line(l)).collect();
+
+    // Second pass: pair openers with closers using a stack, recording
+    // (opener_idx, closer_idx) pairs plus the max fence length found between
+    // them.
+    struct StackEntry {
+        line_idx: usize,
+        fence: FenceLine,
+    }
+
+    let mut stack: Vec<StackEntry> = Vec::new();
+    // Paired blocks: (opener_line, closer_line, max_inner_fence_len)
+    let mut pairs: Vec<(usize, usize, usize)> = Vec::new();
+
+    for (i, fi) in fence_info.iter().enumerate() {
+        let Some(fl) = fi else { continue };
+
+        if fl.has_info {
+            // Labeled fence ⇒ always an opener.
+            stack.push(StackEntry {
+                line_idx: i,
+                fence: fl.clone(),
+            });
+        } else {
+            // Bare fence ⇒ try to close the top of the stack if compatible.
+            let closes_top = stack
+                .last()
+                .is_some_and(|top| top.fence.char == fl.char && fl.len >= top.fence.len);
+            if closes_top {
+                let opener = stack.pop().unwrap();
+                // Find max fence length of any fence line strictly between
+                // opener and closer (these are the nested fences).
+                let inner_max = fence_info[opener.line_idx + 1..i]
+                    .iter()
+                    .filter_map(|fi| fi.as_ref().map(|f| f.len))
+                    .max()
+                    .unwrap_or(0);
+                pairs.push((opener.line_idx, i, inner_max));
+            } else {
+                // Treat as opener.
+                stack.push(StackEntry {
+                    line_idx: i,
+                    fence: fl.clone(),
+                });
+            }
+        }
+    }
+
+    // Determine which lines need rewriting.  A pair needs rewriting when
+    // its opener length <= max inner fence length.
+    struct Rewrite {
+        char: char,
+        new_len: usize,
+        indent: usize,
+    }
+    let mut rewrites: std::collections::HashMap<usize, Rewrite> = std::collections::HashMap::new();
+
+    for (opener_idx, closer_idx, inner_max) in &pairs {
+        let opener_fl = fence_info[*opener_idx].as_ref().unwrap();
+        if opener_fl.len <= *inner_max {
+            let new_len = inner_max + 1;
+            let info_part = {
+                let trimmed = lines[*opener_idx]
+                    .trim_end_matches('\n')
+                    .trim_end_matches('\r');
+                let rest = &trimmed[opener_fl.indent..];
+                rest[opener_fl.len..].to_string()
+            };
+            rewrites.insert(
+                *opener_idx,
+                Rewrite {
+                    char: opener_fl.char,
+                    new_len,
+                    indent: opener_fl.indent,
+                },
+            );
+            let closer_fl = fence_info[*closer_idx].as_ref().unwrap();
+            rewrites.insert(
+                *closer_idx,
+                Rewrite {
+                    char: closer_fl.char,
+                    new_len,
+                    indent: closer_fl.indent,
+                },
+            );
+            // Store info string only in the opener; closer keeps the trailing
+            // portion which is already handled through the original line.
+            // Actually, we rebuild both lines from scratch below, including
+            // the info string for the opener.
+            let _ = info_part; // consumed in rebuild
+        }
+    }
+
+    if rewrites.is_empty() {
+        return markdown.to_string();
+    }
+
+    // Rebuild.
+    let mut out = String::with_capacity(markdown.len() + rewrites.len() * 4);
+    for (i, line) in lines.iter().enumerate() {
+        if let Some(rw) = rewrites.get(&i) {
+            let fence_str: String = std::iter::repeat(rw.char).take(rw.new_len).collect();
+            let indent_str: String = std::iter::repeat(' ').take(rw.indent).collect();
+            // Recover the original info string (if any) and trailing newline.
+            let trimmed = line.trim_end_matches('\n').trim_end_matches('\r');
+            let fi = fence_info[i].as_ref().unwrap();
+            let info = &trimmed[fi.indent + fi.len..];
+            let trailing = &line[trimmed.len()..];
+            out.push_str(&indent_str);
+            out.push_str(&fence_str);
+            out.push_str(info);
+            out.push_str(trailing);
+        } else {
+            out.push_str(line);
+        }
+    }
+    out
+}
+
+fn find_stream_safe_boundary(markdown: &str) -> Option<usize> {
+    let mut open_fence: Option<FenceMarker> = None;
+    let mut last_boundary = None;
+
+    for (offset, line) in markdown.split_inclusive('\n').scan(0usize, |cursor, line| {
+        let start = *cursor;
+        *cursor += line.len();
+        Some((start, line))
+    }) {
+        let line_without_newline = line.trim_end_matches('\n');
+        if let Some(opener) = open_fence {
+            if line_closes_fence(line_without_newline, opener) {
+                open_fence = None;
+                last_boundary = Some(offset + line.len());
+            }
+            continue;
+        }
+
+        if let Some(opener) = parse_fence_opener(line_without_newline) {
+            open_fence = Some(opener);
+            continue;
+        }
+
+        if line_without_newline.trim().is_empty() {
+            last_boundary = Some(offset + line.len());
+        }
+    }
+
+    last_boundary
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+struct FenceMarker {
+    character: char,
+    length: usize,
+}
+
+fn parse_fence_opener(line: &str) -> Option<FenceMarker> {
+    let indent = line.chars().take_while(|c| *c == ' ').count();
+    if indent > 3 {
+        return None;
+    }
+    let rest = &line[indent..];
+    let character = rest.chars().next()?;
+    if character != '`' && character != '~' {
+        return None;
+    }
+    let length = rest.chars().take_while(|c| *c == character).count();
+    if length < 3 {
+        return None;
+    }
+    let info_string = &rest[length..];
+    if character == '`' && info_string.contains('`') {
+        return None;
+    }
+    Some(FenceMarker { character, length })
+}
+
+fn line_closes_fence(line: &str, opener: FenceMarker) -> bool {
+    let indent = line.chars().take_while(|c| *c == ' ').count();
+    if indent > 3 {
+        return false;
+    }
+    let rest = &line[indent..];
+    let length = rest.chars().take_while(|c| *c == opener.character).count();
+    if length < opener.length {
+        return false;
+    }
+    rest[length..].chars().all(|c| c == ' ' || c == '\t')
+}
+
 fn visible_width(input: &str) -> usize {
    strip_ansi(input).chars().count()
 }
@@ -569,7 +906,7 @@ fn strip_ansi(input: &str) -> String {

 #[cfg(test)]
 mod tests {
-    use super::{strip_ansi, Spinner, TerminalRenderer};
+    use super::{strip_ansi, MarkdownStreamState, Spinner, TerminalRenderer};

    #[test]
    fn renders_markdown_with_styling_and_lists() {
@@ -583,16 +920,28 @@ mod tests {
        assert!(markdown_output.contains('\u{1b}'));
    }

+    #[test]
+    fn renders_links_as_colored_markdown_labels() {
+        let terminal_renderer = TerminalRenderer::new();
+        let markdown_output =
+            terminal_renderer.render_markdown("See [Claw](https://example.com/docs) now.");
+        let plain_text = strip_ansi(&markdown_output);
+
+        assert!(plain_text.contains("[Claw](https://example.com/docs)"));
+        assert!(markdown_output.contains('\u{1b}'));
+    }
+
    #[test]
    fn highlights_fenced_code_blocks() {
        let terminal_renderer = TerminalRenderer::new();
        let markdown_output =
-            terminal_renderer.render_markdown("```rust\nfn hi() { println!(\"hi\"); }\n```");
+            terminal_renderer.markdown_to_ansi("```rust\nfn hi() { println!(\"hi\"); }\n```");
        let plain_text = strip_ansi(&markdown_output);

        assert!(plain_text.contains("╭─ rust"));
        assert!(plain_text.contains("fn hi"));
        assert!(markdown_output.contains('\u{1b}'));
+        assert!(markdown_output.contains("[48;5;236m"));
    }

    #[test]
@@ -623,6 +972,80 @@ mod tests {
        assert!(markdown_output.contains('\u{1b}'));
    }

+    #[test]
+    fn streaming_state_waits_for_complete_blocks() {
+        let renderer = TerminalRenderer::new();
+        let mut state = MarkdownStreamState::default();
+
+        assert_eq!(state.push(&renderer, "# Heading"), None);
+        let flushed = state
+            .push(&renderer, "\n\nParagraph\n\n")
+            .expect("completed block");
+        let plain_text = strip_ansi(&flushed);
+        assert!(plain_text.contains("Heading"));
+        assert!(plain_text.contains("Paragraph"));
+
+        assert_eq!(state.push(&renderer, "```rust\nfn main() {}\n"), None);
+        let code = state
+            .push(&renderer, "```\n")
+            .expect("closed code fence flushes");
+        assert!(strip_ansi(&code).contains("fn main()"));
+    }
+
+    #[test]
+    fn streaming_state_holds_outer_fence_with_nested_inner_fence() {
+        let renderer = TerminalRenderer::new();
+        let mut state = MarkdownStreamState::default();
+
+        assert_eq!(
+            state.push(&renderer, "````markdown\n```rust\nfn inner() {}\n"),
+            None,
+            "inner triple backticks must not close the outer four-backtick fence"
+        );
+        assert_eq!(
+            state.push(&renderer, "```\n"),
+            None,
+            "closing the inner fence must not flush the outer fence"
+        );
+        let flushed = state
+            .push(&renderer, "````\n")
+            .expect("closing the outer four-backtick fence flushes the buffered block");
+        let plain_text = strip_ansi(&flushed);
+        assert!(plain_text.contains("fn inner()"));
+        assert!(plain_text.contains("```rust"));
+    }
+
+    #[test]
+    fn streaming_state_distinguishes_backtick_and_tilde_fences() {
+        let renderer = TerminalRenderer::new();
+        let mut state = MarkdownStreamState::default();
+
+        assert_eq!(state.push(&renderer, "~~~text\n"), None);
+        assert_eq!(
+            state.push(&renderer, "```\nstill inside tilde fence\n"),
+            None,
+            "a backtick fence cannot close a tilde-opened fence"
+        );
+        assert_eq!(state.push(&renderer, "```\n"), None);
+        let flushed = state
+            .push(&renderer, "~~~\n")
+            .expect("matching tilde marker closes the fence");
+        let plain_text = strip_ansi(&flushed);
+        assert!(plain_text.contains("still inside tilde fence"));
+    }
+
+    #[test]
+    fn renders_nested_fenced_code_block_preserves_inner_markers() {
+        let terminal_renderer = TerminalRenderer::new();
+        let markdown_output =
+            terminal_renderer.markdown_to_ansi("````markdown\n```rust\nfn nested() {}\n```\n````");
+        let plain_text = strip_ansi(&markdown_output);
+
+        assert!(plain_text.contains("╭─ markdown"));
+        assert!(plain_text.contains("```rust"));
+        assert!(plain_text.contains("fn nested()"));
+    }
+
    #[test]
    fn spinner_advances_frames() {
        let terminal_renderer = TerminalRenderer::new();
@@ -0,0 +1,298 @@
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::process::{Command, Output};
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::time::{SystemTime, UNIX_EPOCH};
+
+use runtime::Session;
+
+static TEMP_COUNTER: AtomicU64 = AtomicU64::new(0);
+
+#[test]
+fn status_command_applies_model_and_permission_mode_flags() {
+    // given
+    let temp_dir = unique_temp_dir("status-flags");
+    fs::create_dir_all(&temp_dir).expect("temp dir should exist");
+
+    // when
+    let output = Command::new(env!("CARGO_BIN_EXE_claw"))
+        .current_dir(&temp_dir)
+        .args([
+            "--model",
+            "sonnet",
+            "--permission-mode",
+            "read-only",
+            "status",
+        ])
+        .output()
+        .expect("claw should launch");
+
+    // then
+    assert_success(&output);
+    let stdout = String::from_utf8(output.stdout).expect("stdout should be utf8");
+    assert!(stdout.contains("Status"));
+    assert!(stdout.contains("Model            claude-sonnet-4-6"));
+    assert!(stdout.contains("Permission mode  read-only"));
+
+    fs::remove_dir_all(temp_dir).expect("cleanup temp dir");
+}
+
+#[test]
+fn resume_flag_loads_a_saved_session_and_dispatches_status() {
+    // given
+    let temp_dir = unique_temp_dir("resume-status");
+    fs::create_dir_all(&temp_dir).expect("temp dir should exist");
+    let session_path = write_session(&temp_dir, "resume-status");
+
+    // when
+    let output = Command::new(env!("CARGO_BIN_EXE_claw"))
+        .current_dir(&temp_dir)
+        .args([
+            "--resume",
+            session_path.to_str().expect("utf8 path"),
+            "/status",
+        ])
+        .output()
+        .expect("claw should launch");
+
+    // then
+    assert_success(&output);
+    let stdout = String::from_utf8(output.stdout).expect("stdout should be utf8");
+    assert!(stdout.contains("Status"));
+    assert!(stdout.contains("Messages         1"));
+    assert!(stdout.contains("Session          "));
+    assert!(stdout.contains(session_path.to_str().expect("utf8 path")));
+
+    fs::remove_dir_all(temp_dir).expect("cleanup temp dir");
+}
+
+#[test]
+fn slash_command_names_match_known_commands_and_suggest_nearby_unknown_ones() {
+    // given
+    let temp_dir = unique_temp_dir("slash-dispatch");
+    fs::create_dir_all(&temp_dir).expect("temp dir should exist");
+
+    // when
+    let help_output = Command::new(env!("CARGO_BIN_EXE_claw"))
+        .current_dir(&temp_dir)
+        .arg("/help")
+        .output()
+        .expect("claw should launch");
+    let unknown_output = Command::new(env!("CARGO_BIN_EXE_claw"))
+        .current_dir(&temp_dir)
+        .arg("/zstats")
+        .output()
+        .expect("claw should launch");
+
+    // then
+    assert_success(&help_output);
+    let help_stdout = String::from_utf8(help_output.stdout).expect("stdout should be utf8");
+    assert!(help_stdout.contains("Interactive slash commands:"));
+    assert!(help_stdout.contains("/status"));
+
+    assert!(
+        !unknown_output.status.success(),
+        "stdout:\n{}\n\nstderr:\n{}",
+        String::from_utf8_lossy(&unknown_output.stdout),
+        String::from_utf8_lossy(&unknown_output.stderr)
+    );
+    let stderr = String::from_utf8(unknown_output.stderr).expect("stderr should be utf8");
+    assert!(stderr.contains("unknown slash command outside the REPL: /zstats"));
+    assert!(stderr.contains("Did you mean"));
+    assert!(stderr.contains("/status"));
+
+    fs::remove_dir_all(temp_dir).expect("cleanup temp dir");
+}
+
+#[test]
+fn omc_namespaced_slash_commands_surface_a_targeted_compatibility_hint() {
+    let temp_dir = unique_temp_dir("slash-dispatch-omc");
+    fs::create_dir_all(&temp_dir).expect("temp dir should exist");
+
+    let output = Command::new(env!("CARGO_BIN_EXE_claw"))
+        .current_dir(&temp_dir)
+        .arg("/oh-my-claudecode:hud")
+        .output()
+        .expect("claw should launch");
+
+    assert!(
+        !output.status.success(),
+        "stdout:\n{}\n\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr)
+    );
+    let stderr = String::from_utf8(output.stderr).expect("stderr should be utf8");
+    assert!(stderr.contains("unknown slash command outside the REPL: /oh-my-claudecode:hud"));
+    assert!(stderr.contains("Claude Code/OMC plugin command"));
+    assert!(stderr.contains("does not yet load plugin slash commands"));
+
+    fs::remove_dir_all(temp_dir).expect("cleanup temp dir");
+}
+
+#[test]
+fn config_command_loads_defaults_from_standard_config_locations() {
+    // given
+    let temp_dir = unique_temp_dir("config-defaults");
+    let config_home = temp_dir.join("home").join(".claw");
+    fs::create_dir_all(temp_dir.join(".claw")).expect("project config dir should exist");
+    fs::create_dir_all(&config_home).expect("home config dir should exist");
+
+    fs::write(config_home.join("settings.json"), r#"{"model":"haiku"}"#)
+        .expect("write user settings");
+    fs::write(temp_dir.join(".claw.json"), r#"{"model":"sonnet"}"#)
+        .expect("write project settings");
+    fs::write(
+        temp_dir.join(".claw").join("settings.local.json"),
+        r#"{"model":"opus"}"#,
+    )
+    .expect("write local settings");
+    let session_path = write_session(&temp_dir, "config-defaults");
+
+    // when
+    let output = command_in(&temp_dir)
+        .env("CLAW_CONFIG_HOME", &config_home)
+        .args([
+            "--resume",
+            session_path.to_str().expect("utf8 path"),
+            "/config",
+            "model",
+        ])
+        .output()
+        .expect("claw should launch");
+
+    // then
+    assert_success(&output);
+    let stdout = String::from_utf8(output.stdout).expect("stdout should be utf8");
+    assert!(stdout.contains("Config"));
+    assert!(stdout.contains("Loaded files      3"));
+    assert!(stdout.contains("Merged section: model"));
+    assert!(stdout.contains("opus"));
+    assert!(stdout.contains(
+        config_home
+            .join("settings.json")
+            .to_str()
+            .expect("utf8 path")
+    ));
+    assert!(stdout.contains(temp_dir.join(".claw.json").to_str().expect("utf8 path")));
+    assert!(stdout.contains(
+        temp_dir
+            .join(".claw")
+            .join("settings.local.json")
+            .to_str()
+            .expect("utf8 path")
+    ));
+
+    fs::remove_dir_all(temp_dir).expect("cleanup temp dir");
+}
+
+#[test]
+fn doctor_command_runs_as_a_local_shell_entrypoint() {
+    // given
+    let temp_dir = unique_temp_dir("doctor-entrypoint");
+    let config_home = temp_dir.join("home").join(".claw");
+    fs::create_dir_all(&config_home).expect("config home should exist");
+
+    // when
+    let output = command_in(&temp_dir)
+        .env("CLAW_CONFIG_HOME", &config_home)
+        .env_remove("ANTHROPIC_API_KEY")
+        .env_remove("ANTHROPIC_AUTH_TOKEN")
+        .env("ANTHROPIC_BASE_URL", "http://127.0.0.1:9")
+        .arg("doctor")
+        .output()
+        .expect("claw doctor should launch");
+
+    // then
+    assert_success(&output);
+    let stdout = String::from_utf8(output.stdout).expect("stdout should be utf8");
+    assert!(stdout.contains("Doctor"));
+    assert!(stdout.contains("Auth"));
+    assert!(stdout.contains("Config"));
+    assert!(stdout.contains("Workspace"));
+    assert!(stdout.contains("Sandbox"));
+    assert!(!stdout.contains("Thinking"));
+
+    fs::remove_dir_all(temp_dir).expect("cleanup temp dir");
+}
+
+#[test]
+fn local_subcommand_help_does_not_fall_through_to_runtime_or_provider_calls() {
+    let temp_dir = unique_temp_dir("subcommand-help");
+    let config_home = temp_dir.join("home").join(".claw");
+    fs::create_dir_all(&config_home).expect("config home should exist");
+
+    let doctor_help = command_in(&temp_dir)
+        .env("CLAW_CONFIG_HOME", &config_home)
+        .env_remove("ANTHROPIC_API_KEY")
+        .env_remove("ANTHROPIC_AUTH_TOKEN")
+        .env("ANTHROPIC_BASE_URL", "http://127.0.0.1:9")
+        .args(["doctor", "--help"])
+        .output()
+        .expect("doctor help should launch");
+    let status_help = command_in(&temp_dir)
+        .env("CLAW_CONFIG_HOME", &config_home)
+        .env_remove("ANTHROPIC_API_KEY")
+        .env_remove("ANTHROPIC_AUTH_TOKEN")
+        .env("ANTHROPIC_BASE_URL", "http://127.0.0.1:9")
+        .args(["status", "--help"])
+        .output()
+        .expect("status help should launch");
+
+    assert_success(&doctor_help);
+    let doctor_stdout = String::from_utf8(doctor_help.stdout).expect("stdout should be utf8");
+    assert!(doctor_stdout.contains("Usage            claw doctor"));
+    assert!(doctor_stdout.contains("local-only health report"));
+    assert!(!doctor_stdout.contains("Thinking"));
+
+    assert_success(&status_help);
+    let status_stdout = String::from_utf8(status_help.stdout).expect("stdout should be utf8");
+    assert!(status_stdout.contains("Usage            claw status"));
+    assert!(status_stdout.contains("local workspace snapshot"));
+    assert!(!status_stdout.contains("Thinking"));
+
+    let doctor_stderr = String::from_utf8(doctor_help.stderr).expect("stderr should be utf8");
+    let status_stderr = String::from_utf8(status_help.stderr).expect("stderr should be utf8");
+    assert!(!doctor_stderr.contains("auth_unavailable"));
+    assert!(!status_stderr.contains("auth_unavailable"));
+
+    fs::remove_dir_all(temp_dir).expect("cleanup temp dir");
+}
+
+fn command_in(cwd: &Path) -> Command {
+    let mut command = Command::new(env!("CARGO_BIN_EXE_claw"));
+    command.current_dir(cwd);
+    command
+}
+
+fn write_session(root: &Path, label: &str) -> PathBuf {
+    let session_path = root.join(format!("{label}.jsonl"));
+    let mut session = Session::new();
+    session
+        .push_user_text(format!("session fixture for {label}"))
+        .expect("session write should succeed");
+    session
+        .save_to_path(&session_path)
+        .expect("session should persist");
+    session_path
+}
+
+fn assert_success(output: &Output) {
+    assert!(
+        output.status.success(),
+        "stdout:\n{}\n\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr)
+    );
+}
+
+fn unique_temp_dir(label: &str) -> PathBuf {
+    let millis = SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .expect("clock should be after epoch")
+        .as_millis();
+    let counter = TEMP_COUNTER.fetch_add(1, Ordering::Relaxed);
+    std::env::temp_dir().join(format!(
+        "claw-{label}-{}-{millis}-{counter}",
+        std::process::id()
+    ))
+}
@@ -0,0 +1,159 @@
+use std::fs;
+use std::path::PathBuf;
+use std::process::{Command, Output};
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::time::{SystemTime, UNIX_EPOCH};
+
+use mock_anthropic_service::{MockAnthropicService, SCENARIO_PREFIX};
+
+static TEMP_COUNTER: AtomicU64 = AtomicU64::new(0);
+
+#[test]
+fn compact_flag_prints_only_final_assistant_text_without_tool_call_details() {
+    // given a workspace pointed at the mock Anthropic service and a fixture file
+    // that the read_file_roundtrip scenario will fetch through a tool call
+    let runtime = tokio::runtime::Runtime::new().expect("tokio runtime should build");
+    let server = runtime
+        .block_on(MockAnthropicService::spawn())
+        .expect("mock service should start");
+    let base_url = server.base_url();
+
+    let workspace = unique_temp_dir("compact-read-file");
+    let config_home = workspace.join("config-home");
+    let home = workspace.join("home");
+    fs::create_dir_all(&workspace).expect("workspace should exist");
+    fs::create_dir_all(&config_home).expect("config home should exist");
+    fs::create_dir_all(&home).expect("home should exist");
+    fs::write(workspace.join("fixture.txt"), "alpha parity line\n").expect("fixture should write");
+
+    // when we run claw in compact text mode against a tool-using scenario
+    let prompt = format!("{SCENARIO_PREFIX}read_file_roundtrip");
+    let output = run_claw(
+        &workspace,
+        &config_home,
+        &home,
+        &base_url,
+        &[
+            "--model",
+            "sonnet",
+            "--permission-mode",
+            "read-only",
+            "--allowedTools",
+            "read_file",
+            "--compact",
+            &prompt,
+        ],
+    );
+
+    // then the command exits successfully and stdout contains exactly the final
+    // assistant text with no tool call IDs, JSON envelopes, or spinner output
+    assert!(
+        output.status.success(),
+        "compact run should succeed\nstdout:\n{}\n\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr),
+    );
+    let stdout = String::from_utf8(output.stdout).expect("stdout should be utf8");
+    let trimmed = stdout.trim_end_matches('\n');
+    assert_eq!(
+        trimmed, "read_file roundtrip complete: alpha parity line",
+        "compact stdout should contain only the final assistant text"
+    );
+    assert!(
+        !stdout.contains("toolu_"),
+        "compact stdout must not leak tool_use_id ({stdout:?})"
+    );
+    assert!(
+        !stdout.contains("\"tool_uses\""),
+        "compact stdout must not leak json envelopes ({stdout:?})"
+    );
+    assert!(
+        !stdout.contains("Thinking"),
+        "compact stdout must not include the spinner banner ({stdout:?})"
+    );
+
+    fs::remove_dir_all(&workspace).expect("workspace cleanup should succeed");
+}
+
+#[test]
+fn compact_flag_streaming_text_only_emits_final_message_text() {
+    // given a workspace pointed at the mock Anthropic service running the
+    // streaming_text scenario which only emits a single assistant text block
+    let runtime = tokio::runtime::Runtime::new().expect("tokio runtime should build");
+    let server = runtime
+        .block_on(MockAnthropicService::spawn())
+        .expect("mock service should start");
+    let base_url = server.base_url();
+
+    let workspace = unique_temp_dir("compact-streaming-text");
+    let config_home = workspace.join("config-home");
+    let home = workspace.join("home");
+    fs::create_dir_all(&workspace).expect("workspace should exist");
+    fs::create_dir_all(&config_home).expect("config home should exist");
+    fs::create_dir_all(&home).expect("home should exist");
+
+    // when we invoke claw with --compact for the streaming text scenario
+    let prompt = format!("{SCENARIO_PREFIX}streaming_text");
+    let output = run_claw(
+        &workspace,
+        &config_home,
+        &home,
+        &base_url,
+        &[
+            "--model",
+            "sonnet",
+            "--permission-mode",
+            "read-only",
+            "--compact",
+            &prompt,
+        ],
+    );
+
+    // then stdout should be exactly the assistant text followed by a newline
+    assert!(
+        output.status.success(),
+        "compact streaming run should succeed\nstdout:\n{}\n\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr),
+    );
+    let stdout = String::from_utf8(output.stdout).expect("stdout should be utf8");
+    assert_eq!(
+        stdout, "Mock streaming says hello from the parity harness.\n",
+        "compact streaming stdout should contain only the final assistant text"
+    );
+
+    fs::remove_dir_all(&workspace).expect("workspace cleanup should succeed");
+}
+
+fn run_claw(
+    cwd: &std::path::Path,
+    config_home: &std::path::Path,
+    home: &std::path::Path,
+    base_url: &str,
+    args: &[&str],
+) -> Output {
+    let mut command = Command::new(env!("CARGO_BIN_EXE_claw"));
+    command
+        .current_dir(cwd)
+        .env_clear()
+        .env("ANTHROPIC_API_KEY", "test-compact-key")
+        .env("ANTHROPIC_BASE_URL", base_url)
+        .env("CLAW_CONFIG_HOME", config_home)
+        .env("HOME", home)
+        .env("NO_COLOR", "1")
+        .env("PATH", "/usr/bin:/bin")
+        .args(args);
+    command.output().expect("claw should launch")
+}
+
+fn unique_temp_dir(label: &str) -> PathBuf {
+    let millis = SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .expect("clock should be after epoch")
+        .as_millis();
+    let counter = TEMP_COUNTER.fetch_add(1, Ordering::Relaxed);
+    std::env::temp_dir().join(format!(
+        "claw-compact-{label}-{}-{millis}-{counter}",
+        std::process::id()
+    ))
+}
@@ -0,0 +1,884 @@
+#![cfg(unix)]
+use std::collections::BTreeMap;
+use std::fs;
+use std::io::Write;
+use std::os::unix::fs::PermissionsExt;
+use std::path::{Path, PathBuf};
+use std::process::{Command, Output, Stdio};
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::time::{SystemTime, UNIX_EPOCH};
+
+use mock_anthropic_service::{MockAnthropicService, SCENARIO_PREFIX};
+use serde_json::{json, Value};
+
+static TEMP_COUNTER: AtomicU64 = AtomicU64::new(0);
+
+#[test]
+#[allow(clippy::too_many_lines)]
+fn clean_env_cli_reaches_mock_anthropic_service_across_scripted_parity_scenarios() {
+    let manifest_entries = load_scenario_manifest();
+    let manifest = manifest_entries
+        .iter()
+        .cloned()
+        .map(|entry| (entry.name.clone(), entry))
+        .collect::<BTreeMap<_, _>>();
+    let runtime = tokio::runtime::Runtime::new().expect("tokio runtime should build");
+    let server = runtime
+        .block_on(MockAnthropicService::spawn())
+        .expect("mock service should start");
+    let base_url = server.base_url();
+
+    let cases = [
+        ScenarioCase {
+            name: "streaming_text",
+            permission_mode: "read-only",
+            allowed_tools: None,
+            stdin: None,
+            prepare: prepare_noop,
+            assert: assert_streaming_text,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "read_file_roundtrip",
+            permission_mode: "read-only",
+            allowed_tools: Some("read_file"),
+            stdin: None,
+            prepare: prepare_read_fixture,
+            assert: assert_read_file_roundtrip,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "grep_chunk_assembly",
+            permission_mode: "read-only",
+            allowed_tools: Some("grep_search"),
+            stdin: None,
+            prepare: prepare_grep_fixture,
+            assert: assert_grep_chunk_assembly,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "write_file_allowed",
+            permission_mode: "workspace-write",
+            allowed_tools: Some("write_file"),
+            stdin: None,
+            prepare: prepare_noop,
+            assert: assert_write_file_allowed,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "write_file_denied",
+            permission_mode: "read-only",
+            allowed_tools: Some("write_file"),
+            stdin: None,
+            prepare: prepare_noop,
+            assert: assert_write_file_denied,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "multi_tool_turn_roundtrip",
+            permission_mode: "read-only",
+            allowed_tools: Some("read_file,grep_search"),
+            stdin: None,
+            prepare: prepare_multi_tool_fixture,
+            assert: assert_multi_tool_turn_roundtrip,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "bash_stdout_roundtrip",
+            permission_mode: "danger-full-access",
+            allowed_tools: Some("bash"),
+            stdin: None,
+            prepare: prepare_noop,
+            assert: assert_bash_stdout_roundtrip,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "bash_permission_prompt_approved",
+            permission_mode: "workspace-write",
+            allowed_tools: Some("bash"),
+            stdin: Some("y\n"),
+            prepare: prepare_noop,
+            assert: assert_bash_permission_prompt_approved,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "bash_permission_prompt_denied",
+            permission_mode: "workspace-write",
+            allowed_tools: Some("bash"),
+            stdin: Some("n\n"),
+            prepare: prepare_noop,
+            assert: assert_bash_permission_prompt_denied,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "plugin_tool_roundtrip",
+            permission_mode: "workspace-write",
+            allowed_tools: None,
+            stdin: None,
+            prepare: prepare_plugin_fixture,
+            assert: assert_plugin_tool_roundtrip,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "auto_compact_triggered",
+            permission_mode: "read-only",
+            allowed_tools: None,
+            stdin: None,
+            prepare: prepare_noop,
+            assert: assert_auto_compact_triggered,
+            extra_env: None,
+            resume_session: None,
+        },
+        ScenarioCase {
+            name: "token_cost_reporting",
+            permission_mode: "read-only",
+            allowed_tools: None,
+            stdin: None,
+            prepare: prepare_noop,
+            assert: assert_token_cost_reporting,
+            extra_env: None,
+            resume_session: None,
+        },
+    ];
+
+    let case_names = cases.iter().map(|case| case.name).collect::<Vec<_>>();
+    let manifest_names = manifest_entries
+        .iter()
+        .map(|entry| entry.name.as_str())
+        .collect::<Vec<_>>();
+    assert_eq!(
+        case_names, manifest_names,
+        "manifest and harness cases must stay aligned"
+    );
+
+    let mut scenario_reports = Vec::new();
+
+    for case in cases {
+        let workspace = HarnessWorkspace::new(unique_temp_dir(case.name));
+        workspace.create().expect("workspace should exist");
+        (case.prepare)(&workspace);
+
+        let run = run_case(case, &workspace, &base_url);
+        (case.assert)(&workspace, &run);
+
+        let manifest_entry = manifest
+            .get(case.name)
+            .unwrap_or_else(|| panic!("missing manifest entry for {}", case.name));
+        scenario_reports.push(build_scenario_report(
+            case.name,
+            manifest_entry,
+            &run.response,
+        ));
+
+        fs::remove_dir_all(&workspace.root).expect("workspace cleanup should succeed");
+    }
+
+    let captured = runtime.block_on(server.captured_requests());
+    // After `be561bf` added count_tokens preflight, each turn sends an
+    // extra POST to `/v1/messages/count_tokens` before the messages POST.
+    // The original count (21) assumed messages-only requests.  We now
+    // filter to `/v1/messages` and verify that subset matches the original
+    // scenario expectation.
+    let messages_only: Vec<_> = captured
+        .iter()
+        .filter(|r| r.path == "/v1/messages")
+        .collect();
+    assert_eq!(
+        messages_only.len(),
+        21,
+        "twelve scenarios should produce twenty-one /v1/messages requests (total captured: {}, includes count_tokens)",
+        captured.len()
+    );
+    assert!(messages_only.iter().all(|request| request.stream));
+
+    let scenarios = messages_only
+        .iter()
+        .map(|request| request.scenario.as_str())
+        .collect::<Vec<_>>();
+    assert_eq!(
+        scenarios,
+        vec![
+            "streaming_text",
+            "read_file_roundtrip",
+            "read_file_roundtrip",
+            "grep_chunk_assembly",
+            "grep_chunk_assembly",
+            "write_file_allowed",
+            "write_file_allowed",
+            "write_file_denied",
+            "write_file_denied",
+            "multi_tool_turn_roundtrip",
+            "multi_tool_turn_roundtrip",
+            "bash_stdout_roundtrip",
+            "bash_stdout_roundtrip",
+            "bash_permission_prompt_approved",
+            "bash_permission_prompt_approved",
+            "bash_permission_prompt_denied",
+            "bash_permission_prompt_denied",
+            "plugin_tool_roundtrip",
+            "plugin_tool_roundtrip",
+            "auto_compact_triggered",
+            "token_cost_reporting",
+        ]
+    );
+
+    let mut request_counts = BTreeMap::new();
+    for request in &captured {
+        *request_counts
+            .entry(request.scenario.as_str())
+            .or_insert(0_usize) += 1;
+    }
+    for report in &mut scenario_reports {
+        report.request_count = *request_counts
+            .get(report.name.as_str())
+            .unwrap_or_else(|| panic!("missing request count for {}", report.name));
+    }
+
+    maybe_write_report(&scenario_reports);
+}
+
+#[derive(Clone, Copy)]
+struct ScenarioCase {
+    name: &'static str,
+    permission_mode: &'static str,
+    allowed_tools: Option<&'static str>,
+    stdin: Option<&'static str>,
+    prepare: fn(&HarnessWorkspace),
+    assert: fn(&HarnessWorkspace, &ScenarioRun),
+    extra_env: Option<(&'static str, &'static str)>,
+    resume_session: Option<&'static str>,
+}
+
+struct HarnessWorkspace {
+    root: PathBuf,
+    config_home: PathBuf,
+    home: PathBuf,
+}
+
+impl HarnessWorkspace {
+    fn new(root: PathBuf) -> Self {
+        Self {
+            config_home: root.join("config-home"),
+            home: root.join("home"),
+            root,
+        }
+    }
+
+    fn create(&self) -> std::io::Result<()> {
+        fs::create_dir_all(&self.root)?;
+        fs::create_dir_all(&self.config_home)?;
+        fs::create_dir_all(&self.home)?;
+        Ok(())
+    }
+}
+
+struct ScenarioRun {
+    response: Value,
+    stdout: String,
+}
+
+#[derive(Debug, Clone)]
+struct ScenarioManifestEntry {
+    name: String,
+    category: String,
+    description: String,
+    parity_refs: Vec<String>,
+}
+
+#[derive(Debug)]
+struct ScenarioReport {
+    name: String,
+    category: String,
+    description: String,
+    parity_refs: Vec<String>,
+    iterations: u64,
+    request_count: usize,
+    tool_uses: Vec<String>,
+    tool_error_count: usize,
+    final_message: String,
+}
+
+fn run_case(case: ScenarioCase, workspace: &HarnessWorkspace, base_url: &str) -> ScenarioRun {
+    let mut command = Command::new(env!("CARGO_BIN_EXE_claw"));
+    command
+        .current_dir(&workspace.root)
+        .env_clear()
+        .env("ANTHROPIC_API_KEY", "test-parity-key")
+        .env("ANTHROPIC_BASE_URL", base_url)
+        .env("CLAW_CONFIG_HOME", &workspace.config_home)
+        .env("HOME", &workspace.home)
+        .env("NO_COLOR", "1")
+        .env("PATH", "/usr/bin:/bin")
+        .args([
+            "--model",
+            "sonnet",
+            "--permission-mode",
+            case.permission_mode,
+            "--output-format=json",
+        ]);
+
+    if let Some(allowed_tools) = case.allowed_tools {
+        command.args(["--allowedTools", allowed_tools]);
+    }
+    if let Some((key, value)) = case.extra_env {
+        command.env(key, value);
+    }
+    if let Some(session_id) = case.resume_session {
+        command.args(["--resume", session_id]);
+    }
+
+    let prompt = format!("{SCENARIO_PREFIX}{}", case.name);
+    command.arg(prompt);
+
+    let output = if let Some(stdin) = case.stdin {
+        let mut child = command
+            .stdin(Stdio::piped())
+            .stdout(Stdio::piped())
+            .stderr(Stdio::piped())
+            .spawn()
+            .expect("claw should launch");
+        child
+            .stdin
+            .as_mut()
+            .expect("stdin should be piped")
+            .write_all(stdin.as_bytes())
+            .expect("stdin should write");
+        child.wait_with_output().expect("claw should finish")
+    } else {
+        command.output().expect("claw should launch")
+    };
+
+    assert_success(&output);
+    let stdout = String::from_utf8_lossy(&output.stdout).into_owned();
+    ScenarioRun {
+        response: parse_json_output(&stdout),
+        stdout,
+    }
+}
+
+#[allow(dead_code)]
+fn prepare_auto_compact_fixture(workspace: &HarnessWorkspace) {
+    let sessions_dir = workspace.root.join(".claw").join("sessions");
+    fs::create_dir_all(&sessions_dir).expect("sessions dir should exist");
+
+    // Write a pre-seeded session with 6 messages so auto-compact can remove them
+    let session_id = "parity-auto-compact-seed";
+    let session_jsonl = r#"{"type":"session_meta","version":3,"session_id":"parity-auto-compact-seed","created_at_ms":1743724800000,"updated_at_ms":1743724800000}
+{"type":"message","message":{"role":"user","blocks":[{"type":"text","text":"step one of the parity scenario"}]}}
+{"type":"message","message":{"role":"assistant","blocks":[{"type":"text","text":"acknowledged step one"}]}}
+{"type":"message","message":{"role":"user","blocks":[{"type":"text","text":"step two of the parity scenario"}]}}
+{"type":"message","message":{"role":"assistant","blocks":[{"type":"text","text":"acknowledged step two"}]}}
+{"type":"message","message":{"role":"user","blocks":[{"type":"text","text":"step three of the parity scenario"}]}}
+{"type":"message","message":{"role":"assistant","blocks":[{"type":"text","text":"acknowledged step three"}]}}
+"#;
+    fs::write(
+        sessions_dir.join(format!("{session_id}.jsonl")),
+        session_jsonl,
+    )
+    .expect("pre-seeded session should write");
+}
+
+fn prepare_noop(_: &HarnessWorkspace) {}
+
+fn prepare_read_fixture(workspace: &HarnessWorkspace) {
+    fs::write(workspace.root.join("fixture.txt"), "alpha parity line\n")
+        .expect("fixture should write");
+}
+
+fn prepare_grep_fixture(workspace: &HarnessWorkspace) {
+    fs::write(
+        workspace.root.join("fixture.txt"),
+        "alpha parity line\nbeta line\ngamma parity line\n",
+    )
+    .expect("grep fixture should write");
+}
+
+fn prepare_multi_tool_fixture(workspace: &HarnessWorkspace) {
+    fs::write(
+        workspace.root.join("fixture.txt"),
+        "alpha parity line\nbeta line\ngamma parity line\n",
+    )
+    .expect("multi tool fixture should write");
+}
+
+fn prepare_plugin_fixture(workspace: &HarnessWorkspace) {
+    let plugin_root = workspace
+        .root
+        .join("external-plugins")
+        .join("parity-plugin");
+    let tool_dir = plugin_root.join("tools");
+    let manifest_dir = plugin_root.join(".claude-plugin");
+    fs::create_dir_all(&tool_dir).expect("plugin tools dir");
+    fs::create_dir_all(&manifest_dir).expect("plugin manifest dir");
+
+    let script_path = tool_dir.join("echo-json.sh");
+    fs::write(
+        &script_path,
+        "#!/bin/sh\nINPUT=$(cat)\nprintf '{\"plugin\":\"%s\",\"tool\":\"%s\",\"input\":%s}\\n' \"$CLAWD_PLUGIN_ID\" \"$CLAWD_TOOL_NAME\" \"$INPUT\"\n",
+    )
+    .expect("plugin script should write");
+    let mut permissions = fs::metadata(&script_path)
+        .expect("plugin script metadata")
+        .permissions();
+    permissions.set_mode(0o755);
+    fs::set_permissions(&script_path, permissions).expect("plugin script should be executable");
+
+    fs::write(
+        manifest_dir.join("plugin.json"),
+        r#"{
+  "name": "parity-plugin",
+  "version": "1.0.0",
+  "description": "mock parity plugin",
+  "tools": [
+    {
+      "name": "plugin_echo",
+      "description": "Echo JSON input",
+      "inputSchema": {
+        "type": "object",
+        "properties": {
+          "message": { "type": "string" }
+        },
+        "required": ["message"],
+        "additionalProperties": false
+      },
+      "command": "./tools/echo-json.sh",
+      "requiredPermission": "workspace-write"
+    }
+  ]
+}"#,
+    )
+    .expect("plugin manifest should write");
+
+    fs::write(
+        workspace.config_home.join("settings.json"),
+        json!({
+            "enabledPlugins": {
+                "parity-plugin@external": true
+            },
+            "plugins": {
+                "externalDirectories": [plugin_root.parent().expect("plugin parent").display().to_string()]
+            }
+        })
+        .to_string(),
+    )
+    .expect("plugin settings should write");
+}
+
+fn assert_streaming_text(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(
+        run.response["message"],
+        Value::String("Mock streaming says hello from the parity harness.".to_string())
+    );
+    assert_eq!(run.response["iterations"], Value::from(1));
+    assert_eq!(run.response["tool_uses"], Value::Array(Vec::new()));
+    assert_eq!(run.response["tool_results"], Value::Array(Vec::new()));
+}
+
+fn assert_read_file_roundtrip(workspace: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(2));
+    assert_eq!(
+        run.response["tool_uses"][0]["name"],
+        Value::String("read_file".to_string())
+    );
+    assert_eq!(
+        run.response["tool_uses"][0]["input"],
+        Value::String(r#"{"path":"fixture.txt"}"#.to_string())
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("alpha parity line"));
+    let output = run.response["tool_results"][0]["output"]
+        .as_str()
+        .expect("tool output");
+    assert!(output.contains(&workspace.root.join("fixture.txt").display().to_string()));
+    assert!(output.contains("alpha parity line"));
+}
+
+fn assert_grep_chunk_assembly(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(2));
+    assert_eq!(
+        run.response["tool_uses"][0]["name"],
+        Value::String("grep_search".to_string())
+    );
+    assert_eq!(
+        run.response["tool_uses"][0]["input"],
+        Value::String(
+            r#"{"pattern":"parity","path":"fixture.txt","output_mode":"count"}"#.to_string()
+        )
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("2 occurrences"));
+    assert_eq!(
+        run.response["tool_results"][0]["is_error"],
+        Value::Bool(false)
+    );
+}
+
+fn assert_write_file_allowed(workspace: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(2));
+    assert_eq!(
+        run.response["tool_uses"][0]["name"],
+        Value::String("write_file".to_string())
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("generated/output.txt"));
+    let generated = workspace.root.join("generated").join("output.txt");
+    let contents = fs::read_to_string(&generated).expect("generated file should exist");
+    assert_eq!(contents, "created by mock service\n");
+    assert_eq!(
+        run.response["tool_results"][0]["is_error"],
+        Value::Bool(false)
+    );
+}
+
+fn assert_write_file_denied(workspace: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(2));
+    assert_eq!(
+        run.response["tool_uses"][0]["name"],
+        Value::String("write_file".to_string())
+    );
+    let tool_output = run.response["tool_results"][0]["output"]
+        .as_str()
+        .expect("tool output");
+    assert!(tool_output.contains("requires workspace-write permission"));
+    assert_eq!(
+        run.response["tool_results"][0]["is_error"],
+        Value::Bool(true)
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("denied as expected"));
+    assert!(!workspace.root.join("generated").join("denied.txt").exists());
+}
+
+fn assert_multi_tool_turn_roundtrip(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(2));
+    let tool_uses = run.response["tool_uses"]
+        .as_array()
+        .expect("tool uses array");
+    assert_eq!(
+        tool_uses.len(),
+        2,
+        "expected two tool uses in a single turn"
+    );
+    assert_eq!(tool_uses[0]["name"], Value::String("read_file".to_string()));
+    assert_eq!(
+        tool_uses[1]["name"],
+        Value::String("grep_search".to_string())
+    );
+    let tool_results = run.response["tool_results"]
+        .as_array()
+        .expect("tool results array");
+    assert_eq!(
+        tool_results.len(),
+        2,
+        "expected two tool results in a single turn"
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("alpha parity line"));
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("2 occurrences"));
+}
+
+fn assert_bash_stdout_roundtrip(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(2));
+    assert_eq!(
+        run.response["tool_uses"][0]["name"],
+        Value::String("bash".to_string())
+    );
+    let tool_output = run.response["tool_results"][0]["output"]
+        .as_str()
+        .expect("tool output");
+    let parsed: Value = serde_json::from_str(tool_output).expect("bash output json");
+    assert_eq!(
+        parsed["stdout"],
+        Value::String("alpha from bash".to_string())
+    );
+    assert_eq!(
+        run.response["tool_results"][0]["is_error"],
+        Value::Bool(false)
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("alpha from bash"));
+}
+
+fn assert_bash_permission_prompt_approved(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert!(run.stdout.contains("Permission approval required"));
+    assert!(run.stdout.contains("Approve this tool call? [y/N]:"));
+    assert_eq!(run.response["iterations"], Value::from(2));
+    assert_eq!(
+        run.response["tool_results"][0]["is_error"],
+        Value::Bool(false)
+    );
+    let tool_output = run.response["tool_results"][0]["output"]
+        .as_str()
+        .expect("tool output");
+    let parsed: Value = serde_json::from_str(tool_output).expect("bash output json");
+    assert_eq!(
+        parsed["stdout"],
+        Value::String("approved via prompt".to_string())
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("approved and executed"));
+}
+
+fn assert_bash_permission_prompt_denied(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert!(run.stdout.contains("Permission approval required"));
+    assert!(run.stdout.contains("Approve this tool call? [y/N]:"));
+    assert_eq!(run.response["iterations"], Value::from(2));
+    let tool_output = run.response["tool_results"][0]["output"]
+        .as_str()
+        .expect("tool output");
+    assert!(tool_output.contains("denied by user approval prompt"));
+    assert_eq!(
+        run.response["tool_results"][0]["is_error"],
+        Value::Bool(true)
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("denied as expected"));
+}
+
+fn assert_plugin_tool_roundtrip(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(2));
+    assert_eq!(
+        run.response["tool_uses"][0]["name"],
+        Value::String("plugin_echo".to_string())
+    );
+    let tool_output = run.response["tool_results"][0]["output"]
+        .as_str()
+        .expect("tool output");
+    let parsed: Value = serde_json::from_str(tool_output).expect("plugin output json");
+    assert_eq!(
+        parsed["plugin"],
+        Value::String("parity-plugin@external".to_string())
+    );
+    assert_eq!(parsed["tool"], Value::String("plugin_echo".to_string()));
+    assert_eq!(
+        parsed["input"]["message"],
+        Value::String("hello from plugin parity".to_string())
+    );
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("hello from plugin parity"));
+}
+
+fn assert_auto_compact_triggered(_: &HarnessWorkspace, run: &ScenarioRun) {
+    // Validates that the auto_compaction field is present in JSON output (format parity).
+    // Trigger behavior is covered by conversation::tests::auto_compacts_when_cumulative_input_threshold_is_crossed.
+    assert_eq!(run.response["iterations"], Value::from(1));
+    assert_eq!(run.response["tool_uses"], Value::Array(Vec::new()));
+    assert!(
+        run.response["message"]
+            .as_str()
+            .expect("message text")
+            .contains("auto compact parity complete."),
+        "expected auto compact message in response"
+    );
+    // auto_compaction key must be present in JSON (may be null for below-threshold sessions)
+    assert!(
+        run.response
+            .as_object()
+            .expect("response object")
+            .contains_key("auto_compaction"),
+        "auto_compaction key must be present in JSON output"
+    );
+    // Verify input_tokens field reflects the large mock token counts
+    let input_tokens = run.response["usage"]["input_tokens"]
+        .as_u64()
+        .expect("input_tokens should be present");
+    assert!(
+        input_tokens >= 50_000,
+        "input_tokens should reflect mock service value (got {input_tokens})"
+    );
+}
+
+fn assert_token_cost_reporting(_: &HarnessWorkspace, run: &ScenarioRun) {
+    assert_eq!(run.response["iterations"], Value::from(1));
+    assert!(run.response["message"]
+        .as_str()
+        .expect("message text")
+        .contains("token cost reporting parity complete."),);
+    let usage = &run.response["usage"];
+    assert!(
+        usage["input_tokens"].as_u64().unwrap_or(0) > 0,
+        "input_tokens should be non-zero"
+    );
+    assert!(
+        usage["output_tokens"].as_u64().unwrap_or(0) > 0,
+        "output_tokens should be non-zero"
+    );
+    assert!(
+        run.response["estimated_cost"]
+            .as_str()
+            .is_some_and(|cost| cost.starts_with('$')),
+        "estimated_cost should be a dollar-prefixed string"
+    );
+}
+
+fn parse_json_output(stdout: &str) -> Value {
+    if let Some(index) = stdout.rfind("{\"auto_compaction\"") {
+        return serde_json::from_str(&stdout[index..]).unwrap_or_else(|error| {
+            panic!("failed to parse JSON response from stdout: {error}\n{stdout}")
+        });
+    }
+
+    stdout
+        .lines()
+        .rev()
+        .find_map(|line| {
+            let trimmed = line.trim();
+            if trimmed.starts_with('{') && trimmed.ends_with('}') {
+                serde_json::from_str(trimmed).ok()
+            } else {
+                None
+            }
+        })
+        .unwrap_or_else(|| panic!("no JSON response line found in stdout:\n{stdout}"))
+}
+
+fn build_scenario_report(
+    name: &str,
+    manifest_entry: &ScenarioManifestEntry,
+    response: &Value,
+) -> ScenarioReport {
+    ScenarioReport {
+        name: name.to_string(),
+        category: manifest_entry.category.clone(),
+        description: manifest_entry.description.clone(),
+        parity_refs: manifest_entry.parity_refs.clone(),
+        iterations: response["iterations"]
+            .as_u64()
+            .expect("iterations should exist"),
+        request_count: 0,
+        tool_uses: response["tool_uses"]
+            .as_array()
+            .expect("tool uses array")
+            .iter()
+            .filter_map(|value| value["name"].as_str().map(ToOwned::to_owned))
+            .collect(),
+        tool_error_count: response["tool_results"]
+            .as_array()
+            .expect("tool results array")
+            .iter()
+            .filter(|value| value["is_error"].as_bool().unwrap_or(false))
+            .count(),
+        final_message: response["message"]
+            .as_str()
+            .expect("message text")
+            .to_string(),
+    }
+}
+
+fn maybe_write_report(reports: &[ScenarioReport]) {
+    let Some(path) = std::env::var_os("MOCK_PARITY_REPORT_PATH") else {
+        return;
+    };
+
+    let payload = json!({
+        "scenario_count": reports.len(),
+        "request_count": reports.iter().map(|report| report.request_count).sum::<usize>(),
+        "scenarios": reports.iter().map(scenario_report_json).collect::<Vec<_>>(),
+    });
+    fs::write(
+        path,
+        serde_json::to_vec_pretty(&payload).expect("report json should serialize"),
+    )
+    .expect("report should write");
+}
+
+fn load_scenario_manifest() -> Vec<ScenarioManifestEntry> {
+    let manifest_path =
+        Path::new(env!("CARGO_MANIFEST_DIR")).join("../../mock_parity_scenarios.json");
+    let manifest = fs::read_to_string(&manifest_path).expect("scenario manifest should exist");
+    serde_json::from_str::<Vec<Value>>(&manifest)
+        .expect("scenario manifest should parse")
+        .into_iter()
+        .map(|entry| ScenarioManifestEntry {
+            name: entry["name"]
+                .as_str()
+                .expect("scenario name should be a string")
+                .to_string(),
+            category: entry["category"]
+                .as_str()
+                .expect("scenario category should be a string")
+                .to_string(),
+            description: entry["description"]
+                .as_str()
+                .expect("scenario description should be a string")
+                .to_string(),
+            parity_refs: entry["parity_refs"]
+                .as_array()
+                .expect("parity refs should be an array")
+                .iter()
+                .map(|value| {
+                    value
+                        .as_str()
+                        .expect("parity ref should be a string")
+                        .to_string()
+                })
+                .collect(),
+        })
+        .collect()
+}
+
+fn scenario_report_json(report: &ScenarioReport) -> Value {
+    json!({
+        "name": report.name,
+        "category": report.category,
+        "description": report.description,
+        "parity_refs": report.parity_refs,
+        "iterations": report.iterations,
+        "request_count": report.request_count,
+        "tool_uses": report.tool_uses,
+        "tool_error_count": report.tool_error_count,
+        "final_message": report.final_message,
+    })
+}
+
+fn assert_success(output: &Output) {
+    assert!(
+        output.status.success(),
+        "stdout:\n{}\n\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr)
+    );
+}
+
+fn unique_temp_dir(label: &str) -> PathBuf {
+    let millis = SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .expect("clock should be after epoch")
+        .as_millis();
+    let counter = TEMP_COUNTER.fetch_add(1, Ordering::Relaxed);
+    std::env::temp_dir().join(format!(
+        "claw-mock-parity-{label}-{}-{millis}-{counter}",
+        std::process::id()
+    ))
+}
@@ -0,0 +1,429 @@
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::process::{Command, Output};
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::time::{SystemTime, UNIX_EPOCH};
+
+use serde_json::Value;
+
+static TEMP_COUNTER: AtomicU64 = AtomicU64::new(0);
+
+#[test]
+fn help_emits_json_when_requested() {
+    let root = unique_temp_dir("help-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let parsed = assert_json_command(&root, &["--output-format", "json", "help"]);
+    assert_eq!(parsed["kind"], "help");
+    assert!(parsed["message"]
+        .as_str()
+        .expect("help text")
+        .contains("Usage:"));
+}
+
+#[test]
+fn version_emits_json_when_requested() {
+    let root = unique_temp_dir("version-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let parsed = assert_json_command(&root, &["--output-format", "json", "version"]);
+    assert_eq!(parsed["kind"], "version");
+    assert_eq!(parsed["version"], env!("CARGO_PKG_VERSION"));
+}
+
+#[test]
+fn status_and_sandbox_emit_json_when_requested() {
+    let root = unique_temp_dir("status-sandbox-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let status = assert_json_command(&root, &["--output-format", "json", "status"]);
+    assert_eq!(status["kind"], "status");
+    assert!(status["workspace"]["cwd"].as_str().is_some());
+
+    let sandbox = assert_json_command(&root, &["--output-format", "json", "sandbox"]);
+    assert_eq!(sandbox["kind"], "sandbox");
+    assert!(sandbox["filesystem_mode"].as_str().is_some());
+}
+
+#[test]
+fn inventory_commands_emit_structured_json_when_requested() {
+    let root = unique_temp_dir("inventory-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let isolated_home = root.join("home");
+    let isolated_config = root.join("config-home");
+    let isolated_codex = root.join("codex-home");
+    fs::create_dir_all(&isolated_home).expect("isolated home should exist");
+
+    let agents = assert_json_command_with_env(
+        &root,
+        &["--output-format", "json", "agents"],
+        &[
+            ("HOME", isolated_home.to_str().expect("utf8 home")),
+            (
+                "CLAW_CONFIG_HOME",
+                isolated_config.to_str().expect("utf8 config home"),
+            ),
+            (
+                "CODEX_HOME",
+                isolated_codex.to_str().expect("utf8 codex home"),
+            ),
+        ],
+    );
+    assert_eq!(agents["kind"], "agents");
+    assert_eq!(agents["action"], "list");
+    assert_eq!(agents["count"], 0);
+    assert_eq!(agents["summary"]["active"], 0);
+    assert!(agents["agents"]
+        .as_array()
+        .expect("agents array")
+        .is_empty());
+
+    let mcp = assert_json_command(&root, &["--output-format", "json", "mcp"]);
+    assert_eq!(mcp["kind"], "mcp");
+    assert_eq!(mcp["action"], "list");
+
+    let skills = assert_json_command(&root, &["--output-format", "json", "skills"]);
+    assert_eq!(skills["kind"], "skills");
+    assert_eq!(skills["action"], "list");
+}
+
+#[test]
+fn agents_command_emits_structured_agent_entries_when_requested() {
+    let root = unique_temp_dir("agents-json-populated");
+    let workspace = root.join("workspace");
+    let project_agents = workspace.join(".codex").join("agents");
+    let home = root.join("home");
+    let user_agents = home.join(".codex").join("agents");
+    let isolated_config = root.join("config-home");
+    let isolated_codex = root.join("codex-home");
+    fs::create_dir_all(&workspace).expect("workspace should exist");
+    write_agent(
+        &project_agents,
+        "planner",
+        "Project planner",
+        "gpt-5.4",
+        "medium",
+    );
+    write_agent(
+        &project_agents,
+        "verifier",
+        "Verification agent",
+        "gpt-5.4-mini",
+        "high",
+    );
+    write_agent(
+        &user_agents,
+        "planner",
+        "User planner",
+        "gpt-5.4-mini",
+        "high",
+    );
+
+    let parsed = assert_json_command_with_env(
+        &workspace,
+        &["--output-format", "json", "agents"],
+        &[
+            ("HOME", home.to_str().expect("utf8 home")),
+            (
+                "CLAW_CONFIG_HOME",
+                isolated_config.to_str().expect("utf8 config home"),
+            ),
+            (
+                "CODEX_HOME",
+                isolated_codex.to_str().expect("utf8 codex home"),
+            ),
+        ],
+    );
+
+    assert_eq!(parsed["kind"], "agents");
+    assert_eq!(parsed["action"], "list");
+    assert_eq!(parsed["count"], 3);
+    assert_eq!(parsed["summary"]["active"], 2);
+    assert_eq!(parsed["summary"]["shadowed"], 1);
+    assert_eq!(parsed["agents"][0]["name"], "planner");
+    assert_eq!(parsed["agents"][0]["source"]["id"], "project_claw");
+    assert_eq!(parsed["agents"][0]["active"], true);
+    assert_eq!(parsed["agents"][1]["name"], "verifier");
+    assert_eq!(parsed["agents"][2]["name"], "planner");
+    assert_eq!(parsed["agents"][2]["active"], false);
+    assert_eq!(parsed["agents"][2]["shadowed_by"]["id"], "project_claw");
+}
+
+#[test]
+fn bootstrap_and_system_prompt_emit_json_when_requested() {
+    let root = unique_temp_dir("bootstrap-system-prompt-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let plan = assert_json_command(&root, &["--output-format", "json", "bootstrap-plan"]);
+    assert_eq!(plan["kind"], "bootstrap-plan");
+    assert!(plan["phases"].as_array().expect("phases").len() > 1);
+
+    let prompt = assert_json_command(&root, &["--output-format", "json", "system-prompt"]);
+    assert_eq!(prompt["kind"], "system-prompt");
+    assert!(prompt["message"]
+        .as_str()
+        .expect("prompt text")
+        .contains("interactive agent"));
+}
+
+#[test]
+fn dump_manifests_and_init_emit_json_when_requested() {
+    let root = unique_temp_dir("manifest-init-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let upstream = write_upstream_fixture(&root);
+    let manifests = assert_json_command_with_env(
+        &root,
+        &["--output-format", "json", "dump-manifests"],
+        &[(
+            "CLAUDE_CODE_UPSTREAM",
+            upstream.to_str().expect("utf8 upstream"),
+        )],
+    );
+    assert_eq!(manifests["kind"], "dump-manifests");
+    assert_eq!(manifests["commands"], 1);
+    assert_eq!(manifests["tools"], 1);
+
+    let workspace = root.join("workspace");
+    fs::create_dir_all(&workspace).expect("workspace should exist");
+    let init = assert_json_command(&workspace, &["--output-format", "json", "init"]);
+    assert_eq!(init["kind"], "init");
+    assert!(workspace.join("CLAUDE.md").exists());
+}
+
+#[test]
+fn doctor_and_resume_status_emit_json_when_requested() {
+    let root = unique_temp_dir("doctor-resume-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let doctor = assert_json_command(&root, &["--output-format", "json", "doctor"]);
+    assert_eq!(doctor["kind"], "doctor");
+    assert!(doctor["message"].is_string());
+    let summary = doctor["summary"].as_object().expect("doctor summary");
+    assert!(summary["ok"].as_u64().is_some());
+    assert!(summary["warnings"].as_u64().is_some());
+    assert!(summary["failures"].as_u64().is_some());
+
+    let checks = doctor["checks"].as_array().expect("doctor checks");
+    assert_eq!(checks.len(), 5);
+    let check_names = checks
+        .iter()
+        .map(|check| {
+            assert!(check["status"].as_str().is_some());
+            assert!(check["summary"].as_str().is_some());
+            assert!(check["details"].is_array());
+            check["name"].as_str().expect("doctor check name")
+        })
+        .collect::<Vec<_>>();
+    assert_eq!(
+        check_names,
+        vec!["auth", "config", "workspace", "sandbox", "system"]
+    );
+
+    let workspace = checks
+        .iter()
+        .find(|check| check["name"] == "workspace")
+        .expect("workspace check");
+    assert!(workspace["cwd"].as_str().is_some());
+    assert!(workspace["in_git_repo"].is_boolean());
+
+    let sandbox = checks
+        .iter()
+        .find(|check| check["name"] == "sandbox")
+        .expect("sandbox check");
+    assert!(sandbox["filesystem_mode"].as_str().is_some());
+    assert!(sandbox["enabled"].is_boolean());
+    assert!(sandbox["fallback_reason"].is_null() || sandbox["fallback_reason"].is_string());
+
+    let session_path = root.join("session.jsonl");
+    fs::write(
+        &session_path,
+        "{\"type\":\"session_meta\",\"version\":3,\"session_id\":\"resume-json\",\"created_at_ms\":0,\"updated_at_ms\":0}\n{\"type\":\"message\",\"message\":{\"role\":\"user\",\"blocks\":[{\"type\":\"text\",\"text\":\"hello\"}]}}\n",
+    )
+    .expect("session should write");
+    let resumed = assert_json_command(
+        &root,
+        &[
+            "--output-format",
+            "json",
+            "--resume",
+            session_path.to_str().expect("utf8 session path"),
+            "/status",
+        ],
+    );
+    assert_eq!(resumed["kind"], "status");
+    // model is null in resume mode (not known without --model flag)
+    assert!(resumed["model"].is_null());
+    assert_eq!(resumed["usage"]["messages"], 1);
+    assert!(resumed["workspace"]["cwd"].as_str().is_some());
+    assert!(resumed["sandbox"]["filesystem_mode"].as_str().is_some());
+}
+
+#[test]
+fn resumed_inventory_commands_emit_structured_json_when_requested() {
+    let root = unique_temp_dir("resume-inventory-json");
+    let config_home = root.join("config-home");
+    let home = root.join("home");
+    fs::create_dir_all(&config_home).expect("config home should exist");
+    fs::create_dir_all(&home).expect("home should exist");
+
+    let session_path = root.join("session.jsonl");
+    fs::write(
+        &session_path,
+        "{\"type\":\"session_meta\",\"version\":3,\"session_id\":\"resume-inventory-json\",\"created_at_ms\":0,\"updated_at_ms\":0}\n{\"type\":\"message\",\"message\":{\"role\":\"user\",\"blocks\":[{\"type\":\"text\",\"text\":\"inventory\"}]}}\n",
+    )
+    .expect("session should write");
+
+    let mcp = assert_json_command_with_env(
+        &root,
+        &[
+            "--output-format",
+            "json",
+            "--resume",
+            session_path.to_str().expect("utf8 session path"),
+            "/mcp",
+        ],
+        &[
+            (
+                "CLAW_CONFIG_HOME",
+                config_home.to_str().expect("utf8 config home"),
+            ),
+            ("HOME", home.to_str().expect("utf8 home")),
+        ],
+    );
+    assert_eq!(mcp["kind"], "mcp");
+    assert_eq!(mcp["action"], "list");
+    assert!(mcp["servers"].is_array());
+
+    let skills = assert_json_command_with_env(
+        &root,
+        &[
+            "--output-format",
+            "json",
+            "--resume",
+            session_path.to_str().expect("utf8 session path"),
+            "/skills",
+        ],
+        &[
+            (
+                "CLAW_CONFIG_HOME",
+                config_home.to_str().expect("utf8 config home"),
+            ),
+            ("HOME", home.to_str().expect("utf8 home")),
+        ],
+    );
+    assert_eq!(skills["kind"], "skills");
+    assert_eq!(skills["action"], "list");
+    assert!(skills["summary"]["total"].is_number());
+    assert!(skills["skills"].is_array());
+}
+
+#[test]
+fn resumed_version_and_init_emit_structured_json_when_requested() {
+    let root = unique_temp_dir("resume-version-init-json");
+    fs::create_dir_all(&root).expect("temp dir should exist");
+
+    let session_path = root.join("session.jsonl");
+    fs::write(
+        &session_path,
+        "{\"type\":\"session_meta\",\"version\":3,\"session_id\":\"resume-version-init-json\",\"created_at_ms\":0,\"updated_at_ms\":0}\n",
+    )
+    .expect("session should write");
+
+    let version = assert_json_command(
+        &root,
+        &[
+            "--output-format",
+            "json",
+            "--resume",
+            session_path.to_str().expect("utf8 session path"),
+            "/version",
+        ],
+    );
+    assert_eq!(version["kind"], "version");
+    assert_eq!(version["version"], env!("CARGO_PKG_VERSION"));
+
+    let init = assert_json_command(
+        &root,
+        &[
+            "--output-format",
+            "json",
+            "--resume",
+            session_path.to_str().expect("utf8 session path"),
+            "/init",
+        ],
+    );
+    assert_eq!(init["kind"], "init");
+    assert!(root.join("CLAUDE.md").exists());
+}
+
+fn assert_json_command(current_dir: &Path, args: &[&str]) -> Value {
+    assert_json_command_with_env(current_dir, args, &[])
+}
+
+fn assert_json_command_with_env(current_dir: &Path, args: &[&str], envs: &[(&str, &str)]) -> Value {
+    let output = run_claw(current_dir, args, envs);
+    assert!(
+        output.status.success(),
+        "stdout:\n{}\n\nstderr:\n{}",
+        String::from_utf8_lossy(&output.stdout),
+        String::from_utf8_lossy(&output.stderr)
+    );
+    serde_json::from_slice(&output.stdout).expect("stdout should be valid json")
+}
+
+fn run_claw(current_dir: &Path, args: &[&str], envs: &[(&str, &str)]) -> Output {
+    let mut command = Command::new(env!("CARGO_BIN_EXE_claw"));
+    command.current_dir(current_dir).args(args);
+    for (key, value) in envs {
+        command.env(key, value);
+    }
+    command.output().expect("claw should launch")
+}
+
+fn write_upstream_fixture(root: &Path) -> PathBuf {
+    let upstream = root.join("claw-code");
+    let src = upstream.join("src");
+    let entrypoints = src.join("entrypoints");
+    fs::create_dir_all(&entrypoints).expect("upstream entrypoints dir should exist");
+    fs::write(
+        src.join("commands.ts"),
+        "import FooCommand from './commands/foo'\n",
+    )
+    .expect("commands fixture should write");
+    fs::write(
+        src.join("tools.ts"),
+        "import ReadTool from './tools/read'\n",
+    )
+    .expect("tools fixture should write");
+    fs::write(
+        entrypoints.join("cli.tsx"),
+        "if (args[0] === '--version') {}\nstartupProfiler()\n",
+    )
+    .expect("cli fixture should write");
+    upstream
+}
+
+fn write_agent(root: &Path, name: &str, description: &str, model: &str, reasoning: &str) {
+    fs::create_dir_all(root).expect("agent root should exist");
+    fs::write(
+        root.join(format!("{name}.toml")),
+        format!(
+            "name = \"{name}\"\ndescription = \"{description}\"\nmodel = \"{model}\"\nmodel_reasoning_effort = \"{reasoning}\"\n"
+        ),
+    )
+    .expect("agent fixture should write");
+}
+
+fn unique_temp_dir(label: &str) -> PathBuf {
+    let millis = SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .expect("clock should be after epoch")
+        .as_millis();
+    let counter = TEMP_COUNTER.fetch_add(1, Ordering::Relaxed);
+    std::env::temp_dir().join(format!(
+        "claw-output-format-{label}-{}-{millis}-{counter}",
+        std::process::id()
+    ))
+}
--- a/Show More
+++ b/Show More