Claude-Code-Workflow/.claude/skills/team-lifecycle-v2/SKILL.md

name, description, allowed-tools

name	description	allowed-tools
team-lifecycle	Unified team skill for full lifecycle - spec/impl/test. All roles invoke this skill with --role arg for role-specific execution. Triggers on "team lifecycle".	TeamCreate(*), TeamDelete(*), SendMessage(*), TaskCreate(*), TaskUpdate(*), TaskList(*), TaskGet(*), Task(*), AskUserQuestion(*), TodoWrite(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*)

Team Lifecycle

Unified team skill covering specification, implementation, testing, and review. All team members invoke this skill with --role=xxx to route to role-specific execution.

Architecture Overview

┌───────────────────────────────────────────────────┐
│  Skill(skill="team-lifecycle")                      │
│  args="任务描述" 或 args="--role=xxx"               │
└───────────────────┬───────────────────────────────┘
                    │ Role Router
                    │
         ┌──── --role present? ────┐
         │ NO                      │ YES
         ↓                         ↓
  Orchestration Mode         Role Dispatch
  (auto → coordinator)      (route to role.md)
         │
    ┌────┴────┬───────┬───────┬───────┬───────┬───────┬───────┐
    ↓         ↓       ↓       ↓       ↓       ↓       ↓       ↓
┌──────────┐┌───────┐┌──────┐┌──────────┐┌───────┐┌────────┐┌──────┐┌────────┐
│coordinator││analyst││writer││discussant││planner││executor││tester││reviewer│
│ roles/   ││roles/ ││roles/││ roles/   ││roles/ ││ roles/ ││roles/││ roles/ │
└──────────┘└───────┘└──────┘└──────────┘└───────┘└────────┘└──────┘└────────┘

Command Architecture

Each role is organized as a folder with a role.md orchestrator and optional commands/ for delegation:

roles/
├── coordinator/
│   ├── role.md              # Orchestrator (Phase 1/5 inline, Phase 2-4 delegate)
│   └── commands/
│       ├── dispatch.md      # Task chain creation (3 modes)
│       └── monitor.md       # Coordination loop + message routing
├── analyst/
│   ├── role.md
│   └── commands/
├── writer/
│   ├── role.md
│   └── commands/
│       └── generate-doc.md  # Multi-CLI document generation (4 doc types)
├── discussant/
│   ├── role.md
│   └── commands/
│       └── critique.md      # Multi-perspective CLI critique
├── planner/
│   ├── role.md
│   └── commands/
│       └── explore.md       # Multi-angle codebase exploration
├── executor/
│   ├── role.md
│   └── commands/
│       └── implement.md     # Multi-backend code implementation
├── tester/
│   ├── role.md
│   └── commands/
│       └── validate.md      # Test-fix cycle
└── reviewer/
    ├── role.md
    └── commands/
        ├── code-review.md   # 4-dimension code review
        └── spec-quality.md  # 5-dimension spec quality check

Design principle: role.md keeps Phase 1 (Task Discovery) and Phase 5 (Report) inline. Phases 2-4 either stay inline (simple logic) or delegate to commands/*.md via Read("commands/xxx.md") when they involve subagent delegation, CLI fan-out, or complex strategies.

Command files are self-contained: each includes Strategy, Execution Steps, and Error Handling. Any subagent can Read() a command file and execute it independently.

Role Router

Input Parsing

Parse $ARGUMENTS to extract --role:

const args = "$ARGUMENTS"
const roleMatch = args.match(/--role[=\s]+(\w+)/)
const teamName = args.match(/--team[=\s]+([\w-]+)/)?.[1] || "lifecycle"

if (!roleMatch) {
  // No --role: Orchestration Mode → auto route to coordinator
  // See "Orchestration Mode" section below
}

const role = roleMatch ? roleMatch[1] : "coordinator"

Role Dispatch

const VALID_ROLES = {
  "coordinator": { file: "roles/coordinator/role.md", prefix: null },
  "analyst":     { file: "roles/analyst/role.md",     prefix: "RESEARCH" },
  "writer":      { file: "roles/writer/role.md",      prefix: "DRAFT" },
  "discussant":  { file: "roles/discussant/role.md",  prefix: "DISCUSS" },
  "planner":     { file: "roles/planner/role.md",     prefix: "PLAN" },
  "executor":    { file: "roles/executor/role.md",    prefix: "IMPL" },
  "tester":      { file: "roles/tester/role.md",      prefix: "TEST" },
  "reviewer":    { file: "roles/reviewer/role.md",    prefix: ["REVIEW", "QUALITY"] }
}

if (!VALID_ROLES[role]) {
  throw new Error(`Unknown role: ${role}. Available: ${Object.keys(VALID_ROLES).join(', ')}`)
}

// Read and execute role-specific logic
Read(VALID_ROLES[role].file)
// → Execute the 5-phase process defined in that file

Orchestration Mode（无参数触发）

当不带 --role 调用时，自动进入 coordinator 编排模式。用户只需传任务描述即可触发完整流程。

触发方式:

// 用户调用（无 --role）— 自动路由到 coordinator
Skill(skill="team-lifecycle", args="任务描述")

// 等价于
Skill(skill="team-lifecycle", args="--role=coordinator 任务描述")

流程:

if (!roleMatch) {
  // Orchestration Mode: 自动路由到 coordinator
  // coordinator role.md 将执行：
  //   Phase 1: 需求澄清
  //   Phase 2: TeamCreate + spawn 所有 worker agents
  //             每个 agent prompt 中包含 Skill(args="--role=xxx") 回调
  //   Phase 3: 创建任务链
  //   Phase 4: 监控协调循环
  //   Phase 5: 结果汇报

  const role = "coordinator"
  Read(VALID_ROLES[role].file)
}

完整调用链:

用户: Skill(args="任务描述")
  │
  ├─ SKILL.md: 无 --role → Orchestration Mode → 读取 coordinator role.md
  │
  ├─ coordinator Phase 2: TeamCreate + spawn workers
  │   每个 worker prompt 中包含 Skill(args="--role=xxx") 回调
  │
  ├─ coordinator Phase 3: dispatch 任务链
  │
  ├─ worker 收到任务 → Skill(args="--role=xxx") → SKILL.md Role Router → role.md
  │   每个 worker 自动获取:
  │   ├─ 角色定义 (role.md: identity, boundaries, message types)
  │   ├─ 可用命令 (commands/*.md)
  │   └─ 执行逻辑 (5-phase process)
  │
  └─ coordinator Phase 4-5: 监控 → 结果汇报

Available Roles

Role	Task Prefix	Responsibility	Role File
coordinator	N/A	Pipeline orchestration, requirement clarification, task dispatch	roles/coordinator/role.md
analyst	RESEARCH-*	Seed analysis, codebase exploration, context gathering	roles/analyst/role.md
writer	DRAFT-*	Product Brief / PRD / Architecture / Epics generation	roles/writer/role.md
discussant	DISCUSS-*	Multi-perspective critique, consensus building	roles/discussant/role.md
planner	PLAN-*	Multi-angle exploration, structured planning	roles/planner/role.md
executor	IMPL-*	Code implementation following plans	roles/executor/role.md
tester	TEST-*	Adaptive test-fix cycles, quality gates	roles/tester/role.md
reviewer	REVIEW-* + QUALITY-*	Code review + Spec quality validation (auto-switch by prefix)	roles/reviewer/role.md

Shared Infrastructure

Role Isolation Rules

核心原则: 每个角色仅能执行自己职责范围内的工作。

Output Tagging（强制）

所有角色的输出必须带 [role_name] 标识前缀：

// SendMessage — content 和 summary 都必须带标识
SendMessage({
  content: `## [${role}] ...`,
  summary: `[${role}] ...`
})

// team_msg — summary 必须带标识
mcp__ccw-tools__team_msg({
  summary: `[${role}] ...`
})

Coordinator 隔离

允许	禁止
需求澄清 (AskUserQuestion)	❌ 直接编写/修改代码
创建任务链 (TaskCreate)	❌ 调用实现类 subagent (code-developer 等)
分发任务给 worker	❌ 直接执行分析/测试/审查
监控进度 (消息总线)	❌ 绕过 worker 自行完成任务
汇报结果给用户	❌ 修改源代码或产物文件

Worker 隔离

允许	禁止
处理自己前缀的任务	❌ 处理其他角色前缀的任务
SendMessage 给 coordinator	❌ 直接与其他 worker 通信
使用 Toolbox 中声明的工具	❌ 为其他角色创建任务 (TaskCreate)
委派给 commands/ 中的命令	❌ 修改不属于本职责的资源

Message Bus (All Roles)

Every SendMessage before, must call mcp__ccw-tools__team_msg to log:

mcp__ccw-tools__team_msg({
  operation: "log",
  team: teamName,
  from: role,
  to: "coordinator",
  type: "<type>",
  summary: `[${role}] <summary>`,
  ref: "<file_path>"
})

Message types by role:

Role	Types
coordinator	plan_approved, plan_revision, task_unblocked, fix_required, error, shutdown
analyst	research_ready, research_progress, error
writer	draft_ready, draft_revision, impl_progress, error
discussant	discussion_ready, discussion_blocked, impl_progress, error
planner	plan_ready, plan_revision, impl_progress, error
executor	impl_complete, impl_progress, error
tester	test_result, impl_progress, fix_required, error
reviewer	review_result, quality_result, fix_required, error

CLI Fallback

When mcp__ccw-tools__team_msg MCP is unavailable:

Bash(`ccw team log --team "${teamName}" --from "${role}" --to "coordinator" --type "<type>" --summary "[${role}] <summary>" --json`)
Bash(`ccw team list --team "${teamName}" --last 10 --json`)
Bash(`ccw team status --team "${teamName}" --json`)

Task Lifecycle (All Worker Roles)

// Standard task lifecycle every worker role follows
// Phase 1: Discovery
const tasks = TaskList()
const prefixes = Array.isArray(VALID_ROLES[role].prefix) ? VALID_ROLES[role].prefix : [VALID_ROLES[role].prefix]
const myTasks = tasks.filter(t =>
  prefixes.some(p => t.subject.startsWith(`${p}-`)) &&
  t.owner === role &&
  t.status === 'pending' &&
  t.blockedBy.length === 0
)
if (myTasks.length === 0) return // idle
const task = TaskGet({ taskId: myTasks[0].id })
TaskUpdate({ taskId: task.id, status: 'in_progress' })

// Phase 1.5: Resume Artifact Check (防止重复产出)
// 当 session 从暂停恢复时，coordinator 已将 in_progress 任务重置为 pending。
// Worker 在开始工作前，必须检查该任务的输出产物是否已存在。
// 如果产物已存在且内容完整：
//   → 直接跳到 Phase 5 报告完成（避免覆盖上次成果）
// 如果产物存在但不完整（如文件为空或缺少关键 section）：
//   → 正常执行 Phase 2-4（基于已有产物继续，而非从头开始）
// 如果产物不存在：
//   → 正常执行 Phase 2-4
//
// 每个 role 检查自己的输出路径:
//   analyst  → sessionFolder/spec/discovery-context.json
//   writer   → sessionFolder/spec/{product-brief.md | requirements/ | architecture/ | epics/}
//   discussant → sessionFolder/discussions/discuss-NNN-*.md
//   planner  → sessionFolder/plan/plan.json
//   executor → git diff (已提交的代码变更)
//   tester   → test pass rate
//   reviewer → sessionFolder/spec/readiness-report.md (quality) 或 review findings (code)

// Phase 2-4: Role-specific (see roles/{role}/role.md)

// Phase 5: Report + Loop — 所有输出必须带 [role] 标识
mcp__ccw-tools__team_msg({ operation: "log", team: teamName, from: role, to: "coordinator", type: "...", summary: `[${role}] ...` })
SendMessage({ type: "message", recipient: "coordinator", content: `## [${role}] ...`, summary: `[${role}] ...` })
TaskUpdate({ taskId: task.id, status: 'completed' })
// Check for next task → back to Phase 1

Three-Mode Pipeline

Spec-only:
  RESEARCH-001 → DISCUSS-001 → DRAFT-001 → DISCUSS-002
  → DRAFT-002 → DISCUSS-003 → DRAFT-003 → DISCUSS-004
  → DRAFT-004 → DISCUSS-005 → QUALITY-001 → DISCUSS-006

Impl-only:
  PLAN-001 → IMPL-001 → TEST-001 + REVIEW-001

Full-lifecycle:
  [Spec pipeline] → PLAN-001(blockedBy: DISCUSS-006) → IMPL-001 → TEST-001 + REVIEW-001

Unified Session Directory

All session artifacts are stored under a single session folder:

.workflow/.team/TLS-{slug}-{YYYY-MM-DD}/
├── team-session.json           # Session state (status, progress, completed_tasks)
├── spec/                       # Spec artifacts (analyst, writer, reviewer output)
│   ├── spec-config.json
│   ├── discovery-context.json
│   ├── product-brief.md
│   ├── requirements/           # _index.md + REQ-*.md + NFR-*.md
│   ├── architecture/           # _index.md + ADR-*.md
│   ├── epics/                  # _index.md + EPIC-*.md
│   ├── readiness-report.md
│   └── spec-summary.md
├── discussions/                # Discussion records (discussant output)
│   └── discuss-001..006.md
└── plan/                       # Plan artifacts (planner output)
    ├── exploration-{angle}.json
    ├── explorations-manifest.json
    ├── plan.json
    └── .task/
        └── TASK-*.json

Messages remain at .workflow/.team-msg/{team-name}/ (unchanged).

Session Resume

Coordinator supports --resume / --continue flags to resume interrupted sessions:

1. Scans .workflow/.team/TLS-*/team-session.json for status: "active" or "paused"
2. Multiple matches → AskUserQuestion for user selection
3. Audit TaskList — 获取当前所有任务的真实状态
4. Reconcile — 双向同步 session.completed_tasks ↔ TaskList 状态:
   • session 已完成但 TaskList 未标记 → 修正 TaskList 为 completed
   • TaskList 已完成但 session 未记录 → 补录到 session
   • in_progress 状态（暂停中断）→ 重置为 pending
5. Determines remaining pipeline from reconciled state
6. Rebuilds team (TeamCreate + worker spawns for needed roles only)
7. Creates missing tasks with correct blockedBy dependency chain (uses TASK_METADATA lookup)
8. Verifies dependency chain integrity for existing tasks
9. Updates session file with reconciled state + current_phase
10. Kick — 向首个可执行任务的 worker 发送 task_unblocked 消息，打破 resume 死锁
11. Jumps to Phase 4 coordination loop

Coordinator Spawn Template

When coordinator creates teammates, use this pattern:

TeamCreate({ team_name: teamName })

// For each worker role:
Task({
  subagent_type: "general-purpose",
  team_name: teamName,
  name: "<role_name>",
  prompt: `你是 team "${teamName}" 的 <ROLE_NAME_UPPER>.

## ⚠️ 首要指令（MUST）
你的所有工作必须通过调用 Skill 获取角色定义后执行，禁止自行发挥：
Skill(skill="team-lifecycle", args="--role=<role_name>")
此调用会加载你的角色定义（role.md）、可用命令（commands/*.md）和完整执行逻辑。

当前需求: ${taskDescription}
约束: ${constraints}
Session: ${sessionFolder}

## 角色准则（强制）
- 你只能处理 <PREFIX>-* 前缀的任务，不得执行其他角色的工作
- 所有输出（SendMessage、team_msg）必须带 [<role_name>] 标识前缀
- 仅与 coordinator 通信，不得直接联系其他 worker
- 不得使用 TaskCreate 为其他角色创建任务

## 消息总线（必须）
每次 SendMessage 前，先调用 mcp__ccw-tools__team_msg 记录。

## 工作流程（严格按顺序）
1. 调用 Skill(skill="team-lifecycle", args="--role=<role_name>") 获取角色定义和执行逻辑
2. 按 role.md 中的 5-Phase 流程执行（TaskList → 找到 <PREFIX>-* 任务 → 执行 → 汇报）
3. team_msg log + SendMessage 结果给 coordinator（带 [<role_name>] 标识）
4. TaskUpdate completed → 检查下一个任务 → 回到步骤 1`
})

See roles/coordinator/role.md for the full spawn implementation with per-role prompts.

Shared Spec Resources

Writer 和 Reviewer 角色在 spec 模式下使用本 skill 内置的标准和模板（从 spec-generator 复制，独立维护）：

Resource	Path	Usage
Document Standards	specs/document-standards.md	YAML frontmatter、命名规范、内容结构
Quality Gates	specs/quality-gates.md	Per-phase 质量门禁、评分标尺
Product Brief Template	templates/product-brief.md	DRAFT-001 文档生成
Requirements Template	templates/requirements-prd.md	DRAFT-002 文档生成
Architecture Template	templates/architecture-doc.md	DRAFT-003 文档生成
Epics Template	templates/epics-template.md	DRAFT-004 文档生成

Writer 在执行每个 DRAFT-* 任务前 必须先 Read 对应的 template 文件和 document-standards.md。 从 roles/ 子目录引用时路径为 ../../specs/ 和 ../../templates/。

Error Handling

Scenario	Resolution
Unknown --role value	Error with available role list
Missing --role arg	Orchestration Mode → auto route to coordinator
Role file not found	Error with expected path (roles/{name}/role.md)
Command file not found	Fall back to inline execution in role.md
Task prefix conflict	Log warning, proceed