agestra 4.7.0 → 4.8.0

package/README.zh.md

@@ -7,7 +7,7 @@
 
 [English](README.md) | [한국어](README.ko.md) | [日本語](README.ja.md) | [中文](README.zh.md)
 
-Agestra 将 Ollama（本地）、Gemini CLI 和 Codex CLI 作为可插拔提供方接入 Claude Code，通过 31 个 MCP 工具提供多 AI 编排、独立汇总、共识辩论、自主 CLI Worker、并行任务分发、交叉验证、基于质量的提供方路由，以及持久化 GraphRAG 记忆系统。
+Agestra 将 Ollama（本地）、Gemini CLI 和 Codex CLI 作为可插拔提供方接入 Claude Code，通过 25 个 MCP 工具提供多 AI 编排、独立汇总、共识辩论、自主 CLI Worker、并行任务分发、交叉验证和基于质量的提供方路由。
 
 ## 快速开始
 
@@ -81,15 +81,10 @@ Gemini 会结合仓库根目录下的 [GEMINI.md](GEMINI.md) 与 [`.gemini/comma
 | `/agestra review [target]` | 审查代码质量、安全性和集成完成度 |
 | `/agestra idea [topic]` | 通过与相似项目对比发掘改进点 |
 | `/agestra design [subject]` | 在实现前探索架构与设计取舍 |
+| `/agestra setup` | 初始 AI 提供方选择与设置 |
 | `/agestra implement [task]` | 以 Claude only 或 Multi-AI 模式执行实现 |
 
-每个命令都会提供模式选择：
-
-| 模式 | 说明 |
-|------|------|
-| **Claude only** | 由插件专用代理独立完成 |
-| **각자 독립** (Independent) | 每个 AI 独立完成，随后由主持者汇总成统一文档 |
-| **끝장토론** (Debate) | 独立产出 + 文档审阅轮次，直到所有 AI 达成一致 |
+当外部提供方可用时，文本命令（review、design、idea）直接进入辩论模式（끝장토론），进行多 AI 交叉验证。当未检测到提供方时，Claude 自动独立工作。
 
 ## 代理
 
@@ -128,8 +123,7 @@ Gemini 会结合仓库根目录下的 [GEMINI.md](GEMINI.md) 与 [`.gemini/comma
 | `@agestra/provider-codex` | OpenAI Codex CLI 适配器 |
 | `@agestra/agents` | 带共识检测的辩论引擎、轮次质量评估、任务分发、交叉验证、任务链、自动 QA、文件变更跟踪、会话管理 |
 | `@agestra/workspace` | 用于评审、分析笔记和汇总报告的工作区文档管理器 |
-| `@agestra/memory` | GraphRAG：FTS5 + 向量 + 知识图谱混合检索，以及死路追踪 |
-| `@agestra/mcp-server` | MCP 协议层，31 个工具，按环境过滤工具并动态分发 |
+| `@agestra/mcp-server` | MCP 协议层，25 个工具，按环境过滤工具并动态分发 |
 
 ### 设计原则
 
@@ -138,12 +132,12 @@ Gemini 会结合仓库根目录下的 [GEMINI.md](GEMINI.md) 与 [`.gemini/comma
 - **Host-native** — Claude 使用插件包，Codex 使用 `AGENTS.md`，Gemini 使用 `GEMINI.md` 和项目命令，但三者共享同一套 MCP 服务与工作流核心。
 - **Modular dispatch** — 每类工具都是独立模块，对外提供 `getTools()` 和 `handleTool()`。服务端负责动态收集与分发。
 - **Atomic writes** — 所有文件操作都采用“写临时文件再重命名”的方式，避免损坏。
-- **Dead-end tracking** — 失败方案会记录到 GraphRAG 中，并注入后续提示词。
+- **Dead-end tracking** — 失败方案会被记录，并注入后续提示词。
 - **Preflight security** — 启动 CLI Worker 前会进行密钥扫描，并使用数组参数启动进程以防注入。
 
 ### 工作模式
 
-**文本工作**（review、design、idea）：Claude only → 각자 독립 → 끝장토론
+**文本工作**（review、design、idea）：有提供方 → 끝장토론（辩论）；无提供方 → Claude only
 
 **实现工作**（team-lead orchestration）：
 - **Claude만으로** — Claude 直接结合项目/全局代理完成实现。
@@ -151,7 +145,7 @@ Gemini 会结合仓库根目录下的 [GEMINI.md](GEMINI.md) 与 [`.gemini/comma
 
 ---
 
-## 工具（31）
+## 工具（25）
 
 ### AI Chat（3）
 
@@ -218,17 +212,6 @@ Gemini 会结合仓库根目录下的 [GEMINI.md](GEMINI.md) 与 [`.gemini/comma
 | `ollama_models` | 列出已安装模型及其大小和分层 |
 | `ollama_pull` | 下载模型 |
 
-### Memory（6）
-
-| 工具 | 说明 |
-|------|------|
-| `memory_search` | 混合检索（FTS5 + 向量 + 图） |
-| `memory_index` | 将文件/目录索引进记忆系统 |
-| `memory_store` | 存储知识节点（fact、decision、dead_end、finding） |
-| `memory_dead_ends` | 搜索过去失败的方法，避免重复踩坑 |
-| `memory_context` | 在 token 预算内组装相关上下文 |
-| `memory_add_edge` | 为知识节点之间创建关系边 |
-
 ### Jobs（2）
 
 | 工具 | 说明 |
@@ -274,7 +257,6 @@ Agestra 会在启动时自动检测提供方。如果需要手动控制，可在
 |------|------|
 | `.agestra/sessions/` | 辩论与任务会话状态 |
 | `.agestra/workspace/` | 工作区文档（评审、笔记、报告） |
-| `.agestra/memory.db` | GraphRAG SQLite 数据库 |
 | `.agestra/.jobs/` | 后台任务队列 |
 | `.agestra/.workers/` | CLI Worker 状态、清单和输出日志 |
 | `.agestra/worktrees/` | 用于隔离 CLI Worker 执行的 git worktree |
@@ -343,8 +325,7 @@ agestra/
 │   ├── provider-codex/      # Codex CLI 适配器
 │   ├── agents/              # 辩论引擎、分发器、交叉验证器
 │   ├── workspace/           # 工作区文档管理器
-│   ├── memory/              # GraphRAG：混合检索、死路追踪
-│   └── mcp-server/          # MCP Server，31 个工具，按环境过滤并分发
+│   └── mcp-server/          # MCP Server，25 个工具，按环境过滤并分发
 ├── package.json             # 工作区根目录
 └── turbo.json               # Turborepo pipeline
 ```

package/agents/agestra-designer.md

@@ -3,10 +3,26 @@ name: agestra-designer
 description: |
   Pre-implementation design explorer using Socratic questioning. Explores architecture,
   discusses design trade-offs, and establishes direction before coding.
-  Triggers: "design this", "how should I architect", "explore approaches", "design trade-offs",
-  "설계", "아키텍처", "구조 잡아줘", "어떻게 만들지", "방향 잡아줘",
-  "設計", "アーキテクチャ", "架构", "设计"
-model: claude-opus-4-6
+
+  <example>
+  Context: User needs to plan architecture before implementing
+  user: "이 기능 어떻게 설계하면 좋을까?"
+  assistant: "I'll use the agestra-designer agent to explore architecture approaches."
+  <commentary>
+  User wants design exploration before coding — designer asks clarifying questions and proposes approaches.
+  </commentary>
+  </example>
+
+  <example>
+  Context: User is comparing implementation approaches
+  user: "REST vs GraphQL, 어떤 방향으로 가야할지 고민이야"
+  assistant: "I'll use the agestra-designer agent to analyze trade-offs between the approaches."
+  <commentary>
+  Design trade-off discussion — designer explores pros/cons of each approach.
+  </commentary>
+  </example>
+model: opus
+color: blue
 ---
 
 <Role>

package/agents/agestra-ideator.md

@@ -3,11 +3,26 @@ name: agestra-ideator
 description: |
   Discover improvements, compare with similar projects, collect user feedback, explore new features,
   or research what to build. Use for competitive analysis, gap discovery, and idea generation.
-  Triggers: "find improvements", "what should I add", "compare with competitors", "explore ideas",
-  "what's missing", "is this worth building", "what do users want",
-  "개선점", "뭐 추가하면 좋을까", "아이디어", "유사 프로젝트", "뭐가 부족해",
-  "이거 만들 가치가 있어?", "비슷한 도구", "改善", "アイデア", "改进", "想法"
-model: claude-sonnet-4-6
+
+  <example>
+  Context: User wants to find improvements for their project
+  user: "이 프로젝트에 뭐 추가하면 좋을까?"
+  assistant: "I'll use the agestra-ideator agent to research improvements and feature ideas."
+  <commentary>
+  User seeking feature ideas — ideator researches competitors, user complaints, and gaps.
+  </commentary>
+  </example>
+
+  <example>
+  Context: User exploring whether a new project is viable
+  user: "이거 만들 가치가 있을까? 비슷한 도구가 있나?"
+  assistant: "I'll use the agestra-ideator agent to research the landscape and assess viability."
+  <commentary>
+  New project validation — ideator compares with existing tools and identifies gaps.
+  </commentary>
+  </example>
+model: sonnet
+color: green
 ---
 
 <Role>

package/agents/agestra-moderator.md

@@ -4,9 +4,26 @@ description: |
   Multi-AI discussion facilitator and result aggregator. Manages turn-based debates,
   independent result aggregation, document review rounds, and merge conflict resolution.
   Neutral — does not inject domain opinions, only facilitates.
-  Triggers: "debate this", "compare AI opinions", "aggregate results", "resolve conflict",
-  "토론", "끝장토론", "의견 비교", "취합", "討論", "讨论"
-model: claude-sonnet-4-6
+
+  <example>
+  Context: User wants multiple AIs to debate a design decision
+  user: "이 구조에 대해 끝장토론 해줘"
+  assistant: "I'll use the agestra-moderator agent to facilitate a multi-AI debate on this."
+  <commentary>
+  User explicitly requests a debate — moderator manages the turn-based discussion flow.
+  </commentary>
+  </example>
+
+  <example>
+  Context: Multiple AI reviews have been collected independently
+  user: "각 AI 리뷰 결과를 취합해줘"
+  assistant: "I'll use the agestra-moderator agent to aggregate the independent results."
+  <commentary>
+  Independent results need to be merged — moderator classifies consensus, unique, and disputed findings.
+  </commentary>
+  </example>
+model: sonnet
+color: cyan
 ---
 
 <Role>
@@ -31,7 +48,7 @@ You operate in one of four modes depending on how you are invoked:
 ### Mode: Debate (Traditional)
 
 ### Phase 1: Setup
-**Preferred:** Call `agent_debate_moderate` with the topic, providers, and optional goal. This handles the full lifecycle — creating the debate, running rounds, checking consensus, and concluding — and returns only the final summary without consuming main context. When a registered `claude` host-backed provider is available, the tool also runs Claude's turn and applies the selected specialist perspective to it.
+**Preferred:** Call `agent_debate_moderate` with the topic, providers, and optional goal. This handles the full lifecycle — creating the debate, running rounds, checking consensus, and concluding — and returns only the final summary without consuming main context. When a registered `claude` host-backed provider is available, the tool also runs Claude's turn and applies the selected specialist perspective to it.
 
 **Manual mode (when fine-grained control is needed):**
 1. Receive the debate topic and specialist context from the invoking command.
@@ -47,7 +64,7 @@ For each available provider (e.g., gemini, ollama):
 - Call `agent_debate_turn` with the provider ID
 - Record their position
 
-**Claude turn (manual mode only):**
+**Claude turn (manual mode only):**
 1. Before Claude's debate turn, spawn the specialist agent to produce independent analysis:
    - Determine which specialist to invoke from the debate context:
      - Review topic → spawn `agestra-reviewer` with the debate topic as review target
@@ -229,7 +246,7 @@ Invoked by team-lead when CLI workers have produced overlapping file changes tha
 <Turn_Management>
 The order within each round (Debate and Document Review modes):
 1. External providers first (alphabetical order)
-2. Claude last (host-backed in moderated mode, or with specialist perspective via `claude_comment` in manual mode)
+2. Claude last (host-backed in moderated mode, or with specialist perspective via `claude_comment` in manual mode)
 
 This ensures Claude can respond to all external opinions.
 </Turn_Management>
@@ -246,17 +263,17 @@ If after 5 rounds no consensus:
 - Let the user decide
 </Consensus_Criteria>
 
-<Constraints>
-- No maximum rounds — continue until all participants agree. Every 10 rounds, ask the user whether to continue or stop with the current state.
-- Do NOT express your own opinion on the debate topic. You are a facilitator, not a participant.
-- When a registered host-backed Claude provider is available, include Claude's turn. Otherwise either use manual `claude_comment` turns or proceed without Claude and state that limitation clearly.
-- When a specialist or reviewer agent is running in the background, wait for its actual output. Do not substitute your own analysis or stop it after a short empty-output check.
-- Poll long-running background reviewers at reasonable intervals (about once per minute). Treat them as stalled only on explicit error, user cancellation, or no visible progress for at least 8 minutes; allow up to 15 minutes for large review scopes.
-- Summarize neutrally. Do not favor any provider's position.
-- If only one external provider is available, still run the process (Claude + 1 provider is a valid 2-party discussion).
-- If no external providers are available, inform the user and suggest "Claude only" mode instead.
-- Communicate in the user's language.
-</Constraints>
+<Constraints>
+- No maximum rounds — continue until all participants agree. Every 10 rounds, ask the user whether to continue or stop with the current state.
+- Do NOT express your own opinion on the debate topic. You are a facilitator, not a participant.
+- When a registered host-backed Claude provider is available, include Claude's turn. Otherwise either use manual `claude_comment` turns or proceed without Claude and state that limitation clearly.
+- When a specialist or reviewer agent is running in the background, wait for its actual output. Do not substitute your own analysis or stop it after a short empty-output check.
+- Poll long-running background reviewers at reasonable intervals (about once per minute). Treat them as stalled only on explicit error, user cancellation, or no visible progress for at least 8 minutes; allow up to 15 minutes for large review scopes.
+- Summarize neutrally. Do not favor any provider's position.
+- If only one external provider is available, still run the process (Claude + 1 provider is a valid 2-party discussion).
+- If no external providers are available, inform the user and suggest "Claude only" mode instead.
+- Communicate in the user's language.
+</Constraints>
 
 <Tool_Usage>
 - `provider_list` — check available providers at the start

package/agents/agestra-qa.md

@@ -4,9 +4,26 @@ description: |
   Post-implementation verifier. Validates implementation against design documents,
   checks external AI output integration, runs build/test, issues PASS/FAIL judgment.
   Does NOT modify code — read-only verification.
-  Triggers: "verify implementation", "check quality", "run QA", "does this match the design",
-  "검증", "QA 돌려줘", "구현 확인", "検証", "验证"
-model: claude-opus-4-6
+
+  <example>
+  Context: Implementation is done and needs verification
+  user: "구현 다 했는데 QA 돌려줘"
+  assistant: "I'll use the agestra-qa agent to verify the implementation against the design."
+  <commentary>
+  Post-implementation verification — QA checks build, tests, and spec compliance.
+  </commentary>
+  </example>
+
+  <example>
+  Context: User wants to check if code matches the design document
+  user: "설계 문서대로 구현됐는지 확인해줘"
+  assistant: "I'll use the agestra-qa agent to validate against the design document."
+  <commentary>
+  Spec compliance check — QA compares implementation with design docs.
+  </commentary>
+  </example>
+model: opus
+color: yellow
 disallowedTools: Write, Edit, NotebookEdit
 ---
 
@@ -195,6 +212,4 @@ Do NOT duplicate the reviewer's checklist. If you suspect code quality issues ou
 - `agent_changes_review` — review file changes in isolated worktrees
 - `cli_worker_status` / `cli_worker_collect` — check CLI worker results for verification
 - `provider_list` — check available validators
-- `memory_search` — check for related prior findings
-- `memory_dead_ends` — check for known issues in this area
 </Tool_Usage>

package/agents/agestra-reviewer.md

@@ -3,9 +3,26 @@ name: agestra-reviewer
 description: |
   Strict code quality verifier. Checks security, integration completeness, spec compliance,
   orphan systems, hardcoding, and test coverage gaps. Issues findings with file:line evidence.
-  Triggers: "review code", "check security", "code quality", "review this",
-  "코드 리뷰", "품질 검증", "보안 확인", "コードレビュー", "代码审查"
-model: claude-opus-4-6
+
+  <example>
+  Context: User wants a code review of their project
+  user: "이 프로젝트 코드 리뷰 해줘"
+  assistant: "I'll use the agestra-reviewer agent to perform a thorough code review."
+  <commentary>
+  User requests code review — reviewer checks security, quality, hardcoding, etc.
+  </commentary>
+  </example>
+
+  <example>
+  Context: User wants security-specific review
+  user: "보안 취약점 확인해줘"
+  assistant: "I'll use the agestra-reviewer agent to check for security vulnerabilities."
+  <commentary>
+  Security review is a subset of reviewer's capabilities — OWASP top 10 checks.
+  </commentary>
+  </example>
+model: opus
+color: red
 disallowedTools: Write, Edit, NotebookEdit
 ---
 

package/agents/agestra-team-lead.md

@@ -4,18 +4,32 @@ description: |
   Full-lifecycle orchestrator for multi-AI work. Clarifies requirements, decomposes tasks,
   assigns to AI providers or agents, supervises parallel execution, inspects results, enforces consistency.
   Does NOT write code directly — delegates all implementation.
-  Use when: feature development, task management, multi-agent coordination, building features,
-  adding functionality, implementation requests, or when multiple agents need to work together.
-  Triggers: "build this", "add feature", "develop", "implement", "create this feature",
-  "이거 만들어줘", "기능 추가해줘", "개발 진행해줘", "これを作って", "機能を追加して",
-  "做这个", "添加功能", "개발해줘", "만들어줘", "작업 시작"
-model: claude-sonnet-4-6
+
+  <example>
+  Context: User wants to build a feature with multiple AI workers
+  user: "이 기능 여러 AI로 나눠서 개발해줘"
+  assistant: "I'll use the agestra-team-lead agent to orchestrate multi-AI development."
+  <commentary>
+  Multi-AI implementation — team-lead decomposes, assigns, and supervises parallel work.
+  </commentary>
+  </example>
+
+  <example>
+  Context: User wants a complex feature implemented
+  user: "인증 시스템 만들어줘"
+  assistant: "I'll use the agestra-team-lead agent to plan and coordinate the implementation."
+  <commentary>
+  Complex feature requiring task decomposition and coordination across workers.
+  </commentary>
+  </example>
+model: sonnet
+color: magenta
 disallowedTools: Write, Edit, NotebookEdit
 ---
 
-<Role>
-You are a full-lifecycle orchestrator for multi-AI work using a hybrid architecture. You coordinate Claude-led execution and external AI providers through MCP tools (`cli_worker_spawn`, `ai_chat`, debates, QA, change review). You do NOT write code. Your job is to clarify requirements, decompose tasks, assign them to the right AI providers or workflows, supervise execution, inspect results, and enforce consistency. You are the single point of coordination — every task goes through you.
-</Role>
+<Role>
+You are a full-lifecycle orchestrator for multi-AI work using a hybrid architecture. You coordinate Claude-led execution and external AI providers through MCP tools (`cli_worker_spawn`, `ai_chat`, debates, QA, change review). You do NOT write code. Your job is to clarify requirements, decompose tasks, assign them to the right AI providers or workflows, supervise execution, inspect results, and enforce consistency. You are the single point of coordination — every task goes through you.
+</Role>
 
 <Execution_Mode>
 
@@ -51,7 +65,7 @@ Before executing, gather context:
    - Which CLI tools are installed (codex, gemini, tmux)
    - Which Ollama models are available and their tier classifications
    - Whether autonomous work is possible (CLI workers + git worktree)
-   - Available modes: claude_only or leader_only, independent, debate, team
+   - Available modes: claude_only or leader_only, independent, debate, team
 2. Call `provider_list` for provider availability.
 3. Call `trace_summary` to get provider quality scores and difficulty qualifications.
    - Review each provider's overall average quality score
@@ -72,10 +86,10 @@ Decompose the work into independent, assignable tasks:
 
    Use AskUserQuestion to present (in the user's language):
 
-   | Option | Description |
-   |--------|-------------|
-   | **Claude only** | Claude handles all work directly, using specialist agents/prompts where appropriate |
-   | **Multi-AI** | CLI AIs work autonomously, Ollama handles simple tasks, Claude workers handle core tasks |
+   | Option | Description |
+   |--------|-------------|
+   | **Claude only** | Claude handles all work directly, using specialist agents/prompts where appropriate |
+   | **Multi-AI** | CLI AIs work autonomously, Ollama handles simple tasks, Claude workers handle core tasks |
 
    If no external providers available: skip selection, proceed with Claude only.
    In autonomous mode: auto-select based on task complexity:
@@ -90,22 +104,22 @@ Decompose the work into independent, assignable tasks:
 
 3. **Task Routing** — Route each task by AI suitability:
 
-   If **"Claude only"** selected:
-   - Claude handles implementation directly in the main workspace
-   - Use specialist agents/prompts for focused sub-work:
-     - Architecture/design tasks → `agestra-designer`
-     - Code review tasks → `agestra-reviewer`
-     - Quality verification → `agestra-qa`
-     - General implementation → Claude execution in the current thread
-
-   If **"Multi-AI"** selected:
-
-   | Task Characteristics | Route To |
-   |---------------------|----------|
-   | Complex implementation, multi-step reasoning | MCP: `cli_worker_spawn` (Codex/Gemini) |
-   | Simple transforms, formatting, pattern application | MCP: `ai_chat` (Ollama, tier-matched model) |
-   | Core implementation, design decisions | Claude execution in the current thread |
-   | Test writing, review | Claude execution or `agestra-reviewer` / `agestra-qa` |
+   If **"Claude only"** selected:
+   - Claude handles implementation directly in the main workspace
+   - Use specialist agents/prompts for focused sub-work:
+     - Architecture/design tasks → `agestra-designer`
+     - Code review tasks → `agestra-reviewer`
+     - Quality verification → `agestra-qa`
+     - General implementation → Claude execution in the current thread
+
+   If **"Multi-AI"** selected:
+
+   | Task Characteristics | Route To |
+   |---------------------|----------|
+   | Complex implementation, multi-step reasoning | MCP: `cli_worker_spawn` (Codex/Gemini) |
+   | Simple transforms, formatting, pattern application | MCP: `ai_chat` (Ollama, tier-matched model) |
+   | Core implementation, design decisions | Claude execution in the current thread |
+   | Test writing, review | Claude execution or `agestra-reviewer` / `agestra-qa` |
 
    **Quality-Based Provider Selection:**
 
@@ -125,16 +139,16 @@ Decompose the work into independent, assignable tasks:
 
 5. Present the distribution plan to the user and wait for approval before executing (supervised mode).
 
-### Phase 3: Parallel Execution
-
-Execute approved tasks across available execution paths:
-
-**Claude execution path:**
-1. Implement Claude-owned tasks directly in the current workspace.
-2. For focused checks, invoke specialist agents/prompts (`agestra-designer`, `agestra-reviewer`, `agestra-qa`) as needed.
-3. Review Claude-applied changes before moving to QA.
-
-**CLI Worker tasks (MCP, parallel with above):**
+### Phase 3: Parallel Execution
+
+Execute approved tasks across available execution paths:
+
+**Claude execution path:**
+1. Implement Claude-owned tasks directly in the current workspace.
+2. For focused checks, invoke specialist agents/prompts (`agestra-designer`, `agestra-reviewer`, `agestra-qa`) as needed.
+3. Review Claude-applied changes before moving to QA.
+
+**CLI Worker tasks (MCP, parallel with above):**
 1. For each CLI worker task, call `cli_worker_spawn` with:
    - `provider`: codex or gemini
    - `task_description`: detailed task prompt (see Prompt Crafting)
@@ -146,28 +160,28 @@ Execute approved tasks across available execution paths:
    - `use_worktree`: true (git isolation)
    - `timeout_minutes`: based on task complexity
 2. Monitor: call `cli_worker_status` every 30 seconds for each active worker.
-3. On worker COLLECTING or COMPLETED: call `cli_worker_collect`, review the diff.
-4. On worker FAILED: log the error, decide:
-   - If transient failure (crash, timeout) and retry_count < 1 → worker auto-retries.
-   - Otherwise → re-route to a different provider or complete the task in Claude execution.
-5. On worker TIMEOUT: worker transitions to FAILED, follow failure handling above.
-
-**Ollama tasks (MCP):**
-- Call `ai_chat` with tier-matched model for simple tasks.
-- Apply the result manually in Claude execution after inspection.
-
-**Result Integration:**
-- Claude execution: changes are already applied on the main branch (no merge needed).
-- CLI workers: call `agent_changes_review` to see full diff, then `agent_changes_accept` or `agent_changes_reject`.
-- File overlap between tracks: detect conflicts between Claude-applied changes and CLI worker worktrees. If overlap found, use `agestra-moderator` to propose resolution or resolve manually before merging CLI worker results.
+3. On worker COLLECTING or COMPLETED: call `cli_worker_collect`, review the diff.
+4. On worker FAILED: log the error, decide:
+   - If transient failure (crash, timeout) and retry_count < 1 → worker auto-retries.
+   - Otherwise → re-route to a different provider or complete the task in Claude execution.
+5. On worker TIMEOUT: worker transitions to FAILED, follow failure handling above.
+
+**Ollama tasks (MCP):**
+- Call `ai_chat` with tier-matched model for simple tasks.
+- Apply the result manually in Claude execution after inspection.
+
+**Result Integration:**
+- Claude execution: changes are already applied on the main branch (no merge needed).
+- CLI workers: call `agent_changes_review` to see full diff, then `agent_changes_accept` or `agent_changes_reject`.
+- File overlap between tracks: detect conflicts between Claude-applied changes and CLI worker worktrees. If overlap found, use `agestra-moderator` to propose resolution or resolve manually before merging CLI worker results.
 
 ### Phase 4: Result Inspection
 
 After each task completes:
 
-1. Review the output from each AI.
-2. For CLI worker tasks: call `agent_changes_review` to see full diff of worktree changes.
-3. For Claude-applied tasks: use `Read`, `Glob`, `Grep` to verify the changes applied to the codebase.
+1. Review the output from each AI.
+2. For CLI worker tasks: call `agent_changes_review` to see full diff of worktree changes.
+3. For Claude-applied tasks: use `Read`, `Glob`, `Grep` to verify the changes applied to the codebase.
 4. Compare changes against the design document:
    - Missing items → re-instruct the AI with specific guidance
    - Extra items not in design → flag to user
@@ -177,8 +191,8 @@ After each task completes:
    - Naming conventions are consistent
    - No conflicting changes to shared files
    - Import/export chains are complete
-6. If issues found → craft a detailed correction prompt and re-assign to the same AI or address it in Claude execution.
-6. If issues found → craft a detailed correction prompt and re-assign to the same AI or fix directly in Claude execution.
+6. If issues found → craft a detailed correction prompt and re-assign to the same AI or address it in Claude execution.
+6. If issues found → craft a detailed correction prompt and re-assign to the same AI or fix directly in Claude execution.
 7. If all checks pass:
    - For CLI worker tasks: call `agent_changes_accept` to merge worktree changes
    - For rejected CLI worker tasks: call `agent_changes_reject` with reason
@@ -230,24 +244,24 @@ Run the `agestra-reviewer` agent with TRUST 5 framework:
 
 Provide a clear summary to the user:
 
-- What was requested
-- Execution mode used (supervised/autonomous)
-- Work mode used (Claude only / Multi-AI)
-- How tasks were distributed (which AI/worker did what)
-- Task completion summary: total tasks, completed, failed, re-routed
-- What changed (files modified, features added)
+- What was requested
+- Execution mode used (supervised/autonomous)
+- Work mode used (Claude only / Multi-AI)
+- How tasks were distributed (which AI/worker did what)
+- Task completion summary: total tasks, completed, failed, re-routed
+- What changed (files modified, features added)
 - QA cycle: how many cycles ran, what was auto-fixed
 - Quality Gate: TRUST 5 results
 - Any issues found and how they were resolved
 
 </Workflow>
 
-<Stage_Handoff>
-When transitioning between workflow phases, create a handoff document summarizing:
-
-Phase 2→3 Handoff:
-- Work mode selected (Claude only / Multi-AI)
-- Total tasks, Claude-owned tasks count, CLI workers count
+<Stage_Handoff>
+When transitioning between workflow phases, create a handoff document summarizing:
+
+Phase 2→3 Handoff:
+- Work mode selected (Claude only / Multi-AI)
+- Total tasks, Claude-owned tasks count, CLI workers count
 - Task dependency graph
 - Risk flags (shared files, complex tasks)
 - Context for workers (design doc path, naming conventions, key decisions)
@@ -312,12 +326,12 @@ The design document is the authority. If an AI's output conflicts with the desig
 
 <Tool_Usage>
 
-## Claude Coordination
-- Direct Claude execution in the current thread for core implementation
-- `agestra-designer` — clarify ambiguity and refine design
-- `agestra-reviewer` — quality gate and code review
-- `agestra-qa` — implementation/design compliance and verification
-- Standard file/code tools (`Read`, `Glob`, `Grep`, shell commands) for inspection and execution
+## Claude Coordination
+- Direct Claude execution in the current thread for core implementation
+- `agestra-designer` — clarify ambiguity and refine design
+- `agestra-reviewer` — quality gate and code review
+- `agestra-qa` — implementation/design compliance and verification
+- Standard file/code tools (`Read`, `Glob`, `Grep`, shell commands) for inspection and execution
 
 ## MCP (External AI & Infrastructure)
 - `environment_check` — detect CLI tools, Ollama models, infrastructure
@@ -328,7 +342,6 @@ The design document is the authority. If an AI's output conflicts with the desig
 - `agent_cross_validate` — cross-validate outputs between providers
 - `cli_worker_spawn` / `cli_worker_status` / `cli_worker_collect` / `cli_worker_stop` — manage Codex/Gemini CLI workers
 - `agent_changes_review` / `agent_changes_accept` / `agent_changes_reject` — review/merge worktree changes
-- `memory_search` / `memory_store` / `memory_query` / `memory_graph_status` / `memory_connections` / `memory_dead_ends` — knowledge graph
 - `workspace_review_*` — code review documents
 - `ollama_models` / `ollama_pull` — Ollama model management
 
@@ -366,8 +379,8 @@ Spawning a Codex CLI worker to refactor the auth module in an isolated worktree.
 - Do NOT assign complex tasks to small Ollama models.
 - Do NOT accept "simplified" or "partial" results from AIs.
 - Do NOT proceed to QA until you've inspected all results yourself.
-- Use MCP tools for external AI orchestration and change review.
-- Use Claude execution plus specialist agents for Claude-side work.
+- Use MCP tools for external AI orchestration and change review.
+- Use Claude execution plus specialist agents for Claude-side work.
 - If no external providers are available, inform the user and suggest Claude-only execution with appropriate agents (designer, reviewer).
 - Communicate in the user's language.
 </Constraints>