maestro-flow 0.5.2 → 0.5.3

package/.agents/skills/maestro-help/index/catalog.json

@@ -73,7 +73,8 @@
     {"name": "team-review", "category": "team", "description": "代码审查团队: 多维度扫描和漏洞审查", "source": "../team-review/SKILL.md"},
     {"name": "team-tech-debt", "category": "team", "description": "技术债务团队: 债务识别和清理", "source": "../team-tech-debt/SKILL.md"},
     {"name": "team-testing", "category": "team", "description": "测试团队: 测试规划和执行", "source": "../team-testing/SKILL.md"},
-    {"name": "codify-to-knowhow", "category": "knowledge", "description": "Manifest 驱动的知识资产生成器", "source": "../codify-to-knowhow/SKILL.md"}
+    {"name": "codify-to-knowhow", "category": "knowledge", "description": "Manifest 驱动的知识资产生成器", "source": "../codify-to-knowhow/SKILL.md"},
+    {"name": "insight-challenge", "category": "quality", "description": "对抗性审查 — 质疑发现、验证声明、产出结构化裁决", "source": "../insight-challenge/SKILL.md"}
   ],
   "agents": [
     {"name": "workflow-analyzer", "description": "多维度分析 — 证据评分和推荐", "category": "workflow"},
@@ -124,7 +125,7 @@
   ],
   "categories": {
     "commands": ["core", "pipeline", "milestone", "manage", "quality", "spec", "learn", "wiki", "worktree", "team", "ralph", "ui", "tools"],
-    "skills": ["meta", "team", "knowledge"],
+    "skills": ["meta", "team", "knowledge", "quality"],
     "agents": ["workflow", "team", "planning", "cli", "ui"]
   },
   "guide_files": [

package/.agents/skills/maestro-ralph/SKILL.md

@@ -271,8 +271,7 @@ resolve_milestone(phase_number):
 | phase 已存在 + 无任何 artifact | `analyze` |
 | phase 已存在 + 最新 artifact = analyze | `plan` |
 | phase 已存在 + 最新 artifact = plan | `execute` |
-| phase 已存在 + 最新 artifact = execute | `verify` |
-| phase 已存在 + 最新 artifact = verify | → refine from result files |
+| phase 已存在 + 最新 artifact = execute | → refine from post-execute results |
 
 **关键不变量**：artifact 过滤按 `session.phase`，不读 `state.json.current_phase`。`phase_is_new` → 直接 `analyze`。
 
@@ -291,11 +290,13 @@ resolve_milestone(phase_number):
 | `medium` / `small` | analyze-macro → plan --from analyze:{ANL_ID} → execute → ...（跳过 roadmap + analyze-phase） |
 | `unknown` | 默认走 large 路径，post-analyze-scope 决策节点再纠正 |
 
-**Refine from verify results:**
+**Refine from post-execute results:**
+
+在 execute artifact 的 scratch dir 中检查结果文件（verification.json 由 execute 内置 gate 产出）：
 
 | Condition | Position |
 |-----------|----------|
-| verification.json: passed==false or gaps[] | `verify-failed` |
+| 无 verification.json 或 passed==false 或 gaps[] | `execute` (触发 post-execute fix loop) |
 | passed==true, no review.json | `business-test` |
 | review.json: verdict=="BLOCK" | `review-failed` |
 | review.json: verdict!="BLOCK" | `test` |
@@ -306,7 +307,7 @@ resolve_milestone(phase_number):
 
 决定下游质量管线长度。读 `session.quality_mode_override`（CLI 标志 `--quality`），无则按规则推断：
 
-| Condition | Mode | Pipeline (verify 之后) |
+| Condition | Mode | Pipeline (execute 之后) |
 |-----------|------|-------------------------|
 | Has `specs/REQ-*.md` + 当前 phase 业务范围明确 | `full` | business-test → review → test-gen → test |
 | Default | `standard` | review → test-gen (当 coverage<80%) → test |
@@ -337,7 +338,7 @@ options:
   - label: "统一规划 (Recommended)"
     description: "一次性分析+规划整个里程碑所有 phase，analyze/plan 走里程碑级，适合 phase 间关联紧密"
   - label: "独立规划"
-    description: "逐个 phase 走完整生命周期（analyze→plan→execute→verify→...），适合 phase 间独立性高"
+    description: "逐个 phase 走完整生命周期（analyze→plan→execute→...），适合 phase 间独立性高"
 ```
 
 写入 `session.planning_mode`（`"unified"` 或 `"independent"`）。`A_BUILD_STEPS` 据此决定 skill args 是否携带 `{phase}` 占位符。
@@ -371,7 +372,7 @@ narrow → derive defaults from intent + codebase, skip questions.
 { "id": "G1", "goal": "<deliverable>", "boundary": "<in/out note>",
   "done_when": "<objectively checkable condition>",
   "evidence": "verification.json|review.json|uat.md|<test path>",
-  "lifecycle": ["analyze","execute","verify"], "status": "pending" }
+  "lifecycle": ["analyze","execute"], "status": "pending" }
 ```
 `done_when` 必须客观可验证，且引用 ralph 已产出的 artifact；`lifecycle` 字段映射到产出 evidence 的生命周期 stage。
 
@@ -412,7 +413,7 @@ Generate steps from `session.lifecycle_position` to `milestone-complete`.
 3. **quality_mode 过滤**：按 `session.quality_mode` 排除不匹配 stage
 3.5. **grill auto_confirm 跳过**：`auto_confirm == true` 时删除 `grill` stage（grill 为交互式苏格拉底拷问，不支持自动模式）；brainstorm args 不含 `--from grill:*`
 4. **决策节点**：每个 Decision after 非空的 stage 之后插入 `{ decision: "<gate>", retry_count: 0, max_retries: 2, command_scope: null, command_path: null }`
-5. **goal-audit 插入**：`task_decomposition` 存在时，在最后一个 evidence-producing stage（verify/review/test）之后、`milestone-complete` 之前插入 `decision:post-goal-audit`
+5. **goal-audit 插入**：`task_decomposition` 存在时，在最后一个 evidence-producing stage（execute/review/test）之后、`milestone-complete` 之前插入 `decision:post-goal-audit`
 6. **终点硬约束**：chain 以 `milestone-complete` 结尾
 7. **goal_ref 传播**：`task_decomposition` 存在时，每个 step 按 `step.stage ∈ g.lifecycle` 匹配 `step.goal_ref = g.id`（多匹配取字典序最小）；decision 节点不打 goal_ref
 8. **占位符**：independent 保留 `{phase}` `{intent}`；unified 不带 `{phase}`
@@ -560,7 +561,7 @@ Runs only when `task_decomposition` present.
 3. 路径 B（`medium` / `small`）：
    - 删除 `goal-audit` 之前所有未完成的 `roadmap` + `analyze` (phase) step
    - 把下一个未完成的 `plan` step 改为 `maestro-plan --from analyze:{analyze_macro_id}`，去掉 `{phase}`，`source_artifact_ref = analyze:{analyze_macro_id}`
-   - 后续 `execute` / `verify` 等沿用同一 standalone scope（不带 `{phase}`，由 plan 写出的 task 列表驱动）
+   - 后续 `execute` 等沿用同一 standalone scope（不带 `{phase}`，由 plan 写出的 task 列表驱动）
 4. 路径 C（`unknown`）：
    - 非 auto_confirm → ask_user 二选一（large / medium-small）；auto_confirm → 默认 large
 5. Reindex steps，标 decision completed，write status.json
@@ -621,7 +622,7 @@ Runs only when `task_decomposition` present.
     "index": 0,
     "skill": "",                  // 执行 step 有值；decision 节点为空字符串/null
     "args": "",
-    "stage": "",                  // brainstorm|blueprint|init|analyze-macro|roadmap|analyze|plan|execute|verify|...
+    "stage": "",                  // brainstorm|blueprint|init|analyze-macro|roadmap|analyze|plan|execute|...
     "scope": null,                // "phase"|"standalone"|"milestone"|null（plan 等需要）
     "decision": null,             // 非 null → decision 节点（值为 gate 名，如 "post-execute"）；null → 执行 step
     "retry_count": 0,             // decision 节点专用

package/.agents/skills/maestro-ralph-beta/SKILL.md

@@ -301,14 +301,15 @@ resolve_milestone(phase_number):
 | phase 已存在 + 无任何 artifact | `analyze` |
 | phase 已存在 + 最新 artifact = analyze | `plan` |
 | phase 已存在 + 最新 artifact = plan | `execute` |
-| phase 已存在 + 最新 artifact = execute | `verify` |
-| phase 已存在 + 最新 artifact = verify | → refine from result files |
+| phase 已存在 + 最新 artifact = execute | → refine from post-execute results |
 
-**Refine from verify results:**
+**Refine from post-execute results:**
+
+在 execute artifact 的 scratch dir 中检查结果文件（verification.json 由 execute 内置 gate 产出）：
 
 | Condition | Position |
 |-----------|----------|
-| verification.json: passed==false or gaps[] | `verify-failed` |
+| 无 verification.json 或 passed==false 或 gaps[] | `execute` (触发 post-execute fix loop) |
 | passed==true, no review.json | `business-test` |
 | review.json: verdict=="BLOCK" | `review-failed` |
 | review.json: verdict!="BLOCK" | `test` |
@@ -334,7 +335,7 @@ resolve_milestone(phase_number):
 
 读 `session.quality_mode_override`（CLI `--quality`），无则按规则推断：
 
-| Condition | Mode | Pipeline (verify 之后) |
+| Condition | Mode | Pipeline (execute 之后) |
 |-----------|------|-------------------------|
 | Has `specs/REQ-*.md` + 当前 phase 业务范围明确 | `full` | business-test → review → test-gen → test |
 | Default | `standard` | review → test-gen (当 coverage<80%) → test |
@@ -363,7 +364,7 @@ options:
   - label: "统一规划 (Recommended)"
     description: "一次性分析+规划整个里程碑所有 phase，analyze/plan 走里程碑级，适合 phase 间关联紧密"
   - label: "独立规划"
-    description: "逐个 phase 走完整生命周期（analyze→plan→execute→verify→...），适合 phase 间独立性高"
+    description: "逐个 phase 走完整生命周期（analyze→plan→execute→...），适合 phase 间独立性高"
 ```
 
 写入 `session.planning_mode`（`"unified"` 或 `"independent"`）。
@@ -395,7 +396,7 @@ narrow → derive defaults from intent + codebase, skip questions.
 { "id": "G1", "goal": "<deliverable>", "boundary": "<in/out note>",
   "done_when": "<objectively checkable condition>",
   "evidence": "verification.json|review.json|uat.md|<test path>",
-  "lifecycle": ["analyze","execute","verify"], "status": "pending" }
+  "lifecycle": ["analyze","execute"], "status": "pending" }
 ```
 `done_when` 必须客观可验证，且引用 ralph 已产出的 artifact；`lifecycle` 字段映射到产出 evidence 的生命周期 stage。
 
@@ -432,7 +433,7 @@ narrow → derive defaults from intent + codebase, skip questions.
 2. **跳过已完成**：跳过当前 milestone+phase 下已有 completed artifact 的 stage（按 `session.phase` 过滤）；unified 按 milestone 过滤
 3. **quality_mode 过滤**：按 `session.quality_mode` 排除不匹配 stage
 4. **决策节点**：每个 Decision after 非空的 stage 之后插入 `{ decision: "<gate>", retry_count: 0, max_retries: 2, command_scope: null, command_path: null }`
-5. **goal-audit 插入**：`task_decomposition` 存在时，在最后一个 evidence-producing stage（verify/review/test）之后、`milestone-complete` 之前插入 `decision:post-goal-audit`
+5. **goal-audit 插入**：`task_decomposition` 存在时，在最后一个 evidence-producing stage（execute/review/test）之后、`milestone-complete` 之前插入 `decision:post-goal-audit`
 6. **终点硬约束**：chain 以 `milestone-complete` 结尾
 7. **goal_ref 传播**：`task_decomposition` 存在时，每个 step 按 `step.stage ∈ g.lifecycle` 匹配 `step.goal_ref = g.id`（多匹配取字典序最小）；decision 节点不打 goal_ref
 8. **占位符**：independent 保留 `{phase}` `{intent}`；unified 不带 `{phase}`
@@ -659,7 +660,7 @@ Write enriched args back to status.json.
 3. `medium` / `small`：
    - 删除 `goal-audit` 之前未完成的 `roadmap` + `analyze` (phase) step
    - 下一个未完成的 `plan` step → `maestro-plan --from analyze:{analyze_macro_id}`，去掉 `{phase}`，`source_artifact_ref = analyze:{analyze_macro_id}`
-   - 后续 `execute` / `verify` 同 standalone scope
+   - 后续 `execute` 等沿用同一 standalone scope
 4. `unknown`：非 auto_confirm → ask_user 二选一（large / medium-small）；auto_confirm → 默认 large
 5. release 协议 — 完成分支；reindex
 6. Display: ◆ Scope verdict: {verdict} → {kept|collapsed to standalone via analyze:{ANL_ID}}
@@ -756,7 +757,7 @@ Display: `[{index}/{total}] ✗ {step.skill} 失败，会话已暂停。/maestro
     "index": 0,
     "skill": "",                   // 执行 step 有值；decision 节点为空字符串/null
     "args": "",
-    "stage": "",                   // brainstorm|blueprint|init|analyze-macro|roadmap|analyze|plan|execute|verify|...
+    "stage": "",                   // brainstorm|blueprint|init|analyze-macro|roadmap|analyze|plan|execute|...
     "scope": null,                 // "phase"|"standalone"|"milestone"|null（plan 等需要）
     "decision": null,              // null = 执行 step；非 null = decision step (值为 gate 名)
     "retry_count": 0,              // decision step

package/.agents/skills/odyssey-debug/SKILL.md

@@ -0,0 +1,475 @@
+---
+name: odyssey-debug
+description: Long-running debug cycle — archaeology, diagnosis, fix, confirmation, generalization, discovery, and knowledge persistence
+argument-hint: "<issue> [--skip-fix] [--skip-generalize] [--auto] [-y] [-c]"
+allowed-tools:
+  - read_file
+  - write_file
+  - edit_file
+  - shell
+  - find_files
+  - search
+  - delegate_subagent
+  - ask_user
+---
+<!-- Open-standard mirror generated by scripts/build-agents-standard.mjs — do not edit; re-run after editing .claude/ source. -->
+
+<purpose>
+Closed-loop deep debugging: archaeology (what changed) → explore (call chains, error gaps) → diagnose (hypothesis-driven) → fix & confirm → generalize (举一反三) → discover siblings → persist learnings.
+
+Unlike `quality-debug` (fast fix), this treats every bug as a learning signal — digs into git history before hypotheses, confirms fixes with CLI review, scans for siblings of the root cause.
+
+Core philosophy:
+- **Archaeology before hypothesis** — look at what changed before guessing why
+- **Fix one, find many** — a single bug reveals a class of bugs
+- **Decision journal** — human-judgment items recorded, not lost
+- **CLI-assisted review** — delegate for second-opinion analysis
+
+**三句哲学约束（穷尽迭代）:**
+1. **零遗留** — 根因必须确认到底，修复必须验证通过，泛化必须扫描穷尽
+2. **穷尽迭代** — 假设失败不放弃：扩范围 → 换视角 → 升级工具，直到根因确认或明确 INCONCLUSIVE
+3. **改进即标准** — 修复后重新确认同区域无新问题，泛化发现的同类 bug 全部处理
+
+Entry: `/odyssey-debug "issue"` (full cycle) | `-c` (resume) | `--skip-fix` (analysis-only)
+</purpose>
+
+<boundary>
+**范围内:** 单一 bug/issue 的完整闭环 — 考古 → 探索 → 诊断 → 修复 → 确认 → 泛化同类 → 沉淀
+**范围外:** 新功能开发 → `/odyssey-planex` | 代码质量审查 → `/odyssey-review-test-fix` | UI 视觉优化 → `/odyssey-ui` | 架构重设计 → `/maestro-plan`
+**探索自由度:** 边界内自由探索 — 可追踪任意调用链、分析任意历史、测试任意假设。泛化阶段可扫描全项目寻找同类问题。
+**模板支持:** `--template <name>` 从预定义调查策略启动，跳过假设生成直接进入针对性诊断：
+
+| Template | 调查策略 | 适用场景 |
+|----------|---------|---------|
+| `performance` | profiling → hot path → allocation → cache | 性能劣化 |
+| `memory-leak` | heap snapshot → retention chain → lifecycle | 内存泄漏 |
+| `race-condition` | timeline → concurrent access → lock analysis | 竞态条件 |
+| `regression` | git bisect → diff analysis → boundary check | 回归问题 |
+| `crash` | stack trace → null chain → error propagation | 崩溃/异常 |
+</boundary>
+
+<execution_discipline>
+**三条铁律（所有阶段适用）:**
+
+1. **Phase auto-commit** — 每个阶段完成后**自动** `git commit`，无需用户确认
+   - 代码变更 + understanding.md → `git add` → `git commit -m "odyssey-debug({slug}): {phase} — {摘要}"`
+   - session.json / evidence.ndjson 为运行时状态，不纳入 commit
+   - 确保每个阶段的进展可回溯、可恢复
+
+2. **有把握才改** — 仅修改自己有把握的内容；不确定的记录决策等人判断
+   - 有把握 → 直接修改代码，commit
+   - 需要决策 → 记录 `evidence.ndjson {"phase":"decision","status":"pending"}` 不改代码
+   - 禁止猜测性修改，宁可多记录一条 decision 也不冒险改错
+
+3. **多 CLI 辅助** — 利用 `maestro delegate` 调用多个 CLI 工具交叉验证
+   - 关键判断用不同 `--role`（analyze / review / explore）获取多视角
+   - 修复前后各做一次 CLI review 确认
+   - 不同阶段可调用不同工具，综合多方意见再行动
+</execution_discipline>
+
+<context>
+$ARGUMENTS — issue description and optional flags.
+
+**Flags:** `--skip-fix` analysis-only | `--skip-generalize` quick fix | `--template <name>` 预定义策略 | `--auto` no delegate confirmation | `-y` auto-confirm all decisions | `-c` resume last session
+
+**Session**: `SESSION_DIR = .workflow/scratch/{YYYYMMDD}-debug-odyssey-{slug}/`
+
+**Output — 4 files:**
+```
+SESSION_DIR/
+  ├── session.json       # state + confirmation + patterns + phase_goals
+  ├── evidence.ndjson    # append-only evidence trail (phase field distinguishes origin)
+  ├── explore.json       # structured CLI exploration snapshot
+  └── understanding.md   # evolving narrative — 9 sections
+```
+
+**session.json schema:**
+```json
+{
+  "session_id": "debug-odyssey-{YYYYMMDD-HHmmss}", "issue": "",
+  "flags": { "skip_fix": false, "skip_generalize": false, "auto": false, "auto_confirm": false },
+  "current_state": "S_INTAKE", "diagnosis_retries": 0,
+  "root_cause": null, "patterns": [], "confirmation": null,
+  "phase_goals": [], "phase_goals_all_done": false, "self_iteration_log": [],
+  "generalization_stats": null,
+  "cross_phase_loops": 0, "max_loops": 3,
+  "created_at": "", "updated_at": ""
+}
+```
+
+**evidence.ndjson — unified trail:**
+```json
+{"ts":"","phase":"archaeology|explore|diagnosis|discovery|decision|self-iteration","type":"","source":"","content":"","note":""}
+```
+Phase-specific fields:
+- `archaeology`: `sha`, `author`, `date`, `message`, `relevance` (high|medium|low)
+- `explore`: `category` (call_chain|recent_change|error_gap|similar_pattern), `detail`
+- `diagnosis`: `hypothesis`, `result` (confirmed|disproved|inconclusive)
+- `discovery`: `file`, `line`, `classification` (safe|risk|bug), `action` (fix|issue|decision|skip)
+- `decision`: `question`, `options`, `context`, `status` (pending|resolved|deferred), `resolution`
+- `self-iteration`: `stage`, `round`, `assessment`, `expansion`
+
+**explore.json schema:**
+```json
+{
+  "call_chains": [{"entry":"","chain":["file:line"]}],
+  "recent_changes": [{"file":"","commits":[{"sha":"","message":"","date":""}]}],
+  "error_gaps": [{"file":"","line":0,"description":""}],
+  "similar_patterns": [{"file":"","line":0,"description":""}],
+  "cli_tool": "", "timestamp": ""
+}
+```
+
+**phase_goals[] — auto-derived from flags:**
+
+| ID | Goal | done_when | phase | skip_when |
+|----|------|-----------|-------|-----------|
+| G1 | Root cause identified | evidence.ndjson has phase=diagnosis result=confirmed | S_DIAGNOSE | — |
+| G2 | Explore context gathered | explore.json ≥1 category populated | S_EXPLORE | — |
+| G3 | Fix applied and confirmed | confirmation.overall == confirmed | S_CONFIRM | skip_fix |
+| G4 | Pattern generalized | patterns[] ≥1 entry | S_GENERALIZE | skip_generalize |
+| G5 | Discoveries triaged | all scan hits classified | S_DISCOVER | skip_generalize |
+| G6 | Learnings persisted | spec entries created OR no actionable learnings | S_RECORD | — |
+
+When `flags[skip_when] == true` → auto set `status: "skipped"`, `completion_confirmed: true`.
+
+**understanding.md — 9 sections (written by owning phase):**
+1. Issue & Scope ← S_INTAKE | 2. Archaeology Summary ← S_ARCHAEOLOGY | 3. Exploration ← S_EXPLORE
+4. Hypotheses & Testing ← S_DIAGNOSE | 5. Root Cause ← S_DIAGNOSE | 6. Fix & Confirmation ← S_FIX+S_CONFIRM
+7. Generalization ← S_GENERALIZE | 8. Discoveries & Decisions ← S_DISCOVER | 9. Learnings ← S_RECORD
+
+### Pre-load（可选，缺失不阻塞）
+
+| 层级 | 命令 | 作用 |
+|------|------|------|
+| Codebase docs | Read `.workflow/codebase/ARCHITECTURE.md` | 模块边界，作为所有分析的上下文 |
+| Wiki search | `maestro search "<issue keywords>" --json` | 先前调查、相关决策（取 top 5） |
+| Specs + tools | `maestro spec load --category debug --keyword "<symptom>"` | 已知 issue/workaround + 可发现的 knowhow 工具 |
+| Role knowledge | `maestro search --category debug` → 选相关 → `maestro wiki load <id>` | 累积领域知识 |
+| Prior sessions | `find_files(".workflow/scratch/*-debug-odyssey-*")` | 相关 odyssey 会话 |
+
+### Knowledge Persistence（S_RECORD 中写入产出文件）
+
+S_RECORD 阶段将可沉淀知识 **写入 understanding.md §9 Learnings**，按以下分类结构化：
+
+| 分类 | 写入内容 | 后续建议命令 |
+|------|---------|-------------|
+| 反复根因模式 | 模式描述 + 触发条件 + 修复模板 | `/spec-add debug "..."` |
+| 非显而易见 workaround | 问题场景 + 解决方案 + 适用范围 | `/spec-add learning "..."` |
+| 架构边界违反 | 违反描述 + 正确边界 + 检查方法 | `/spec-add arch "..."` |
+| 可复用泛化 pattern | pattern 签名 + 风险说明 + fix 模板 | `/spec-add coding "..."` |
+
+**两步模式：** 执行中写入产出文件（临时记录）→ 任务完成后用户通过 next_step_routing 沉淀为永久知识。执行过程中不调用外部 Skill。
+</context>
+
+<self_iteration>
+**Quality Gate (适用: S_ARCHAEOLOGY, S_EXPLORE, S_DIAGNOSE, S_GENERALIZE)**
+
+| 维度 | sufficient | insufficient |
+|------|-----------|-------------|
+| Coverage | 已知相关文件/模块均已分析 | 遗漏 grep/git log 可发现的目标 |
+| Depth | ≥80% 发现有 file:line 级证据 | 多数仅泛泛描述 |
+| Actionability | 每条结论有具体后续动作 | 仅"建议关注"类无操作性结论 |
+
+**规则:** 阶段完成 → 评估 3 维度 → 任一 insufficient → 重入（每阶段最多 2 轮）。
+- Round 1: 扩范围 — 增加目录、git log depth ×2、增加 delegate 角度
+- Round 2: 换视角 — 不同 CLI tool、反向追踪、手动 code reading
+
+**退出:** 全 sufficient → 推进 | 2 轮上限 → 记录 gap 继续。记录至 `evidence.ndjson` + `session.json.self_iteration_log[]`.
+</self_iteration>
+
+<state_machine>
+
+<states>
+S_INTAKE       — 解析问题、加载上下文、检查/恢复 session     PERSIST: session.json + understanding.md §1
+S_ARCHAEOLOGY  — 考古：git history + CLI 分析              PERSIST: evidence.ndjson (archaeology) + understanding.md §2
+S_EXPLORE      — CLI 探索：调用链、错误间隙、相似模式        PERSIST: explore.json + evidence.ndjson (explore) + understanding.md §3
+S_DIAGNOSE     — 假设驱动根因分析                          PERSIST: evidence.ndjson (diagnosis|decision) + understanding.md §4-5
+S_FIX          — 实现修复 (skip_fix 时跳过)                PERSIST: code changes + evidence.ndjson (decision)
+S_CONFIRM      — 测试 + CLI review 双重确认 (skip_fix 时跳过) PERSIST: session.json.confirmation + understanding.md §6
+S_GENERALIZE   — 举一反三：提取 pattern，扫描相似代码       PERSIST: session.json.patterns + understanding.md §7
+S_DISCOVER     — 评估发现，创建 issue / 记录决策            PERSIST: evidence.ndjson (discovery|decision) + understanding.md §8
+S_RECORD       — 知识沉淀 + 目标审计                       PERSIST: understanding.md §9 + spec entries
+</states>
+
+<transitions>
+S_INTAKE → S_INTAKE       : -c + session found → A_RESUME_SESSION
+S_INTAKE → S_ARCHAEOLOGY  : issue parsed → A_INTAKE
+S_INTAKE → S_INTAKE       : no issue, no session → ask_user
+
+S_ARCHAEOLOGY → S_EXPLORE     : A_ARCHAEOLOGY complete
+S_EXPLORE     → S_DIAGNOSE    : A_EXPLORE complete
+
+S_DIAGNOSE → S_FIX          : root cause confirmed, !skip_fix
+S_DIAGNOSE → S_GENERALIZE   : root cause confirmed, skip_fix, !skip_generalize
+S_DIAGNOSE → S_RECORD       : root cause confirmed, skip_fix, skip_generalize
+S_DIAGNOSE → S_DIAGNOSE     : all hypotheses failed, retries < 3 → A_ESCALATE_DIAGNOSIS
+S_DIAGNOSE → S_RECORD       : all hypotheses failed, retries >= 3 → mark INCONCLUSIVE
+
+S_FIX     → S_CONFIRM       : fix implemented
+S_CONFIRM → S_GENERALIZE    : confirmed, !skip_generalize
+S_CONFIRM → S_RECORD        : confirmed, skip_generalize
+S_CONFIRM → S_FIX           : needs_rework
+
+S_GENERALIZE → S_DISCOVER   : similar code found
+S_GENERALIZE → S_RECORD     : no similar code
+
+S_DISCOVER → S_DIAGNOSE     : discovery finds new bug worth investigating, loops < max_loops → cross_phase_loops++
+S_DISCOVER → S_FIX          : discovery finds same-pattern bug, fix template applies, loops < max_loops → cross_phase_loops++
+S_DISCOVER → S_RECORD       : triage complete OR loops >= max_loops (剩余项 → issue/decision)
+
+S_RECORD   → END            : A_RECORD complete
+</transitions>
+
+<actions>
+
+### A_INTAKE
+1. Parse arguments: issue description, flags
+2. Generate slug, create `SESSION_DIR`
+3. Search: `maestro search "<keywords>"` + Glob prior sessions + ARCHITECTURE.md + Grep keywords
+4. Derive `phase_goals[]` from flags (apply `skip_when`)
+5. Write `session.json` + `understanding.md` §1
+6. Emit Goal Prompt (see Appendix)
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): INTAKE — 目标解析与上下文加载"`
+
+### A_RESUME_SESSION
+Find latest session via Glob → read `session.json` → display summary → jump to `current_state`.
+
+### A_ARCHAEOLOGY
+**Git archaeology (2 parallel Agents):**
+
+| Agent | Task |
+|-------|------|
+| Timeline | `git log --oneline -20 -- {files}` → change timeline |
+| Blame | Top 3 suspicious files: `git blame -L {region}` → critical paths |
+
+Append findings to `evidence.ndjson` (phase: "archaeology").
+
+**CLI-assisted change review:**
+```bash
+maestro delegate "PURPOSE: Review recent modifications related to: {issue}
+TASK: Analyze intent behind changes | Identify risky modifications | Flag potential bug sources
+MODE: analysis
+CONTEXT: @{relevant_files} | Git log: {top_10_commits}
+EXPECTED: JSON [{commit_sha, risk_level, analysis, could_cause_issue, explanation}]
+CONSTRAINTS: Focus on behavioral changes, not formatting
+" --role analyze --mode analysis
+```
+Run_in_background, STOP, wait for callback. Append results to evidence.
+
+Update `understanding.md` §2.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): ARCHAEOLOGY — git 考古分析"`
+
+### A_EXPLORE
+Skip if no enabled CLI tools (W006).
+
+```bash
+maestro delegate "PURPOSE: Gather codebase evidence for: {issue}
+TASK: Trace call chains | Find recent changes | Identify error gaps | Check similar patterns
+MODE: analysis
+CONTEXT: @**/*
+EXPECTED: JSON {call_chains, recent_changes, error_gaps, similar_patterns}
+CONSTRAINTS: Max 20 entries/category | Symptom-related code paths
+Symptoms: {issue}  Archaeology hints: {suspicious_commits}
+" --role explore --mode analysis
+```
+Run_in_background, STOP, wait for callback.
+
+Parse → write `explore.json` + append `evidence.ndjson` (phase: "explore"). Update §3. Mark G2 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): EXPLORE — 代码探索完成"`
+
+### A_DIAGNOSE
+1. **Form hypotheses** from evidence (archaeology + explore), ranked [HIGH]/[MEDIUM]/[LOW] → §4
+2. **Test each** (rank order): design test → execute → append evidence (phase: "diagnosis")
+3. **Decision journal**: ambiguity → evidence (phase: "decision"); Normal: ask_user | `-y`: defer
+4. **Root cause**: confirmed → `session.json.root_cause` + §5. Mark G1 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): DIAGNOSE — 根因确认"`
+
+### A_ESCALATE_DIAGNOSIS
+Increment `diagnosis_retries`. If < 3: broaden scope via `maestro delegate --role analyze` (same delegate format), form new hypotheses, return to S_DIAGNOSE. If >= 3: Normal → ask_user (broaden/new/INCONCLUSIVE) | `-y` → auto INCONCLUSIVE, proceed to S_RECORD. See Appendix: `-y` behavior.
+
+### A_FIX
+1. Present root cause + proposed fix. Normal: ask_user | `-y`: auto proceed (see Appendix)
+2. Implement fix
+3. Record in evidence (phase: "decision")
+
+📌 **Auto-commit**: `git add -A && git commit -m "odyssey-debug({slug}): FIX — {修复摘要}"`
+
+### A_CONFIRM
+1. **Tests**: auto-detect framework, run covering tests
+2. **CLI fix review**:
+```bash
+maestro delegate "PURPOSE: Review fix for: {issue}
+TASK: Verify correctness | Check regressions | Assess completeness | Review edge cases
+MODE: analysis
+CONTEXT: @{modified_files} | Root cause: {summary} | Diff: {git_diff}
+EXPECTED: JSON {verdict, findings [{severity, description, suggestion}], regression_risk}
+CONSTRAINTS: Focus on correctness, not style
+" --role review --mode analysis
+```
+Run_in_background, STOP, wait for callback.
+
+3. Write `session.json.confirmation`: `{test_result, cli_review, overall: "confirmed|needs_rework", timestamp}`
+4. Update §6. `needs_rework` → S_FIX. `confirmed` → mark G3 done, advance.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): CONFIRM — 修复验证"`
+
+### A_GENERALIZE
+举一反三: multi-layer pattern extraction → 4-agent scan → cross-layer dedup → iterative deepening.
+
+**Pattern extraction** from root cause + fix:
+
+| Layer | Method | Example |
+|-------|--------|---------|
+| Syntax | Regex patterns (direct Grep) | `eval(`, missing `await`, unclosed resource |
+| Semantic | Anti-pattern description (Agent-driven) | Unhandled async errors, unvalidated input |
+| Structural | Architecture-level (file/module similarity) | Same import structure, missing override |
+
+Write `session.json.patterns[]`: `[{id, source, layer, signature, description, risk, fix_template}]`
+
+**4-agent parallel codebase scan:**
+
+| Agent | Strategy | Input | Scope |
+|-------|----------|-------|-------|
+| Syntax grep | Grep syntax-layer regex | P*.signature | Full project |
+| Semantic scan | Understand + check anti-pattern | P*.description | Related modules |
+| Structural match | Find structurally similar files | Buggy file structure | Full project |
+| Historical grep | `git log -S "{pattern}"` | P*.signature | Full git history |
+
+Returns: `[{pattern_id, file, line, context, risk_level, layer, confidence}]`
+
+**Cross-layer dedup**: same file:line multi-layer hit → boost confidence | single-layer → `needs_review` | historical hit on fixed record → `regression_risk`
+
+**Iterative deepening**: module with ≥3 hits → targeted deep scan (max 1 round).
+
+**Quality Gate** (self-iteration) → if insufficient, expand and re-scan.
+
+Write §7 + `session.json.generalization_stats`: `{patterns_extracted, total_hits, cross_layer_confirmed, regression_risks, by_layer, deepening_triggered}`. Mark G4 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): GENERALIZE — 泛化扫描完成"`
+
+### A_DISCOVER
+1. **Triage** each hit: read ±10 lines context → classify `safe`/`risk`/`bug`
+2. **Route**: see Appendix `-y` behavior table. Append evidence (phase: "discovery" + "decision")
+3. Update §8. Mark G5 done.
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): DISCOVER — 发现分类完成"`
+
+### A_RECORD
+1. Finalize `understanding.md` §9
+2. **Write learnings** to understanding.md §9: 按 Knowledge Persistence 表分类记录（临时），completion summary 列出建议的 `/spec-add` 命令
+3. Mark G6 done. Process pending decisions: Normal → ask_user | `-y` → skip (show deferred count)
+4. **Goal audit**: all `completion_confirmed` true → `phase_goals_all_done = true`. Any false: Normal → ask_user (回退/跳过/接受) | `-y` → auto accept
+5. **Completion**: `current_state = "COMPLETED"`, emit summary:
+```
+--- DEBUG ODYSSEY COMPLETE ---
+Issue:      {issue}
+Root cause: {root_cause.hypothesis}
+Fix:        {applied|skipped|inconclusive}
+Patterns:   {patterns_extracted} ({by_layer} distribution)
+Scan hits:  {total_hits} ({cross_layer_confirmed} cross-layer confirmed)
+Issues:     {N} created
+Decisions:  {N} resolved, {M} pending, {K} deferred
+Learnings:  {N} spec entries persisted
+Self-iter:  {N} quality gate rounds across {M} stages
+Goals:      {done}/{total} confirmed ({skipped} skipped)
+---
+```
+
+📌 **Auto-commit**: `git add understanding.md && git commit -m "odyssey-debug({slug}): RECORD — 会话总结与知识沉淀"`
+
+</actions>
+
+<appendix>
+
+### Goal Prompt Template
+
+**时机守卫：仅在 A_INTAKE 完成后显示一次。A_RECORD 完成时禁止重新显示。**
+
+```
+📋 Debug Odyssey 会话已创建。可随时复制以下 /goal 设定终止条件：
+
+/goal 穷尽迭代：直到根因确认（或明确 INCONCLUSIVE）且修复验证通过
+且泛化扫描穷尽（所有 hits 已分类处理）且 phase_goals_all_done=true 才停。
+假设失败时扩范围→换视角→升级工具，不轻易放弃。
+泛化发现的同类 bug 全部修复或创建 issue，不允许遗留。
+遇到 phase=decision 的 pending 必须 ask_user，不得自行 resolve。
+```
+
+Odyssey 输出提示词后继续执行不阻塞。`/goal` 由用户任意时刻输入。
+
+### `-y` Auto-Confirm Behavior
+
+| Decision Point | Normal | `-y` mode |
+|---------------|--------|-----------|
+| A_DIAGNOSE ambiguity | ask_user blocks | record `deferred`, best-effort continue |
+| A_ESCALATE 3-strike | ask_user 3-way | auto INCONCLUSIVE |
+| A_FIX direction | ask_user confirm | auto proceed with suggested fix |
+| A_DISCOVER bug triage | ask_user route | auto create issue |
+| A_DISCOVER ambiguous | ask_user batch | all `deferred` |
+| A_RECORD decisions | ask_user per-item | skip, show deferred count |
+| A_RECORD goal audit | ask_user 3-way | auto accept current state |
+
+`deferred` items shown as "待决策" in completion summary; recoverable via `-c`.
+
+### Phase Goal Lifecycle
+
+`pending → done (confirmed=true)` normal | `pending → skipped (confirmed=true)` flags/manual | `pending → failed (confirmed=false)` INCONCLUSIVE
+
+`phase_goals_all_done = true` only when ALL goals have `completion_confirmed == true`.
+
+</appendix>
+
+</state_machine>
+
+<error_codes>
+| Code | Severity | Condition | Recovery |
+|------|----------|-----------|----------|
+| E001 | error | No issue and no session to resume | Provide issue or use -c |
+| E003 | error | Resume but no session found | Start new session |
+| E004 | error | Delegate execution failed | Retry or proceed without CLI |
+| W001 | warning | No relevant git history | Proceed with limited context |
+| W002 | warning | All hypotheses inconclusive after 3 retries | INCONCLUSIVE |
+| W003 | warning | Generalization scan 0 hits | Skip discovery |
+| W004 | warning | Delegate parse failed | Use raw output |
+| W005 | warning | Pending decisions unresolved | Filter evidence.ndjson phase=decision |
+| W006 | warning | CLI exploration skipped (no tools) | Proceed without explore.json |
+</error_codes>
+
+<success_criteria>
+- [ ] Session directory created with 4 output files
+- [ ] Prior knowledge searched (maestro search + sessions + architecture)
+- [ ] Git archaeology (log + blame) + CLI change review, evidence.ndjson phase=archaeology
+- [ ] CLI exploration, explore.json written, evidence phase=explore
+- [ ] Hypotheses formed from archaeology + explore, tested and logged phase=diagnosis
+- [ ] Root cause declared with evidence refs
+- [ ] understanding.md tracks all 9 sections progressively
+- [ ] Fix implemented + confirmed with test + CLI review (unless --skip-fix)
+- [ ] Multi-layer patterns (syntax/semantic/structural) extracted (unless --skip-generalize)
+- [ ] 4-agent scan + cross-layer dedup + iterative deepening for ≥3 hits/module
+- [ ] Discoveries classified and routed (fix/issue/decision/skip)
+- [ ] Decision journal: all human-judgment items in evidence.ndjson phase=decision
+- [ ] phase_goals derived from flags, skip_when applied, each phase marks its goal
+- [ ] Goal audit in A_RECORD — unmet goals surfaced, phase_goals_all_done set correctly
+- [ ] Goal Prompt displayed once after session creation
+- [ ] `-y`: all decisions auto-resolve/defer, deferred count in summary
+- [ ] State saved at each transition (resumable via -c)
+- [ ] Quality Gate self-iteration when insufficient, logged in self_iteration_log
+- [ ] Spec entries persisted for reusable learnings
+- [ ] Completion summary with all stats
+</success_criteria>
+
+<next_step_routing>
+| Condition | Next step |
+|-----------|-----------|
+| Issues from discoveries | `/manage-issue list --source debug-odyssey` |
+| Pattern worth documenting | `/learn-decompose <module>` |
+| Fix needs formal review | `/quality-review <phase>` |
+| Second opinion on root cause | `/learn-second-opinion <understanding.md>` |
+| Related question | `/learn-investigate "<question>"` |
+| Decisions still pending | Filter evidence.ndjson phase=decision status=pending |
+</next_step_routing>