mirror of
https://github.com/catlog22/Claude-Code-Workflow.git
synced 2026-02-28 09:23:08 +08:00
- Added `findAllCoreMemoryDatabases` function to recursively locate core-memory databases in nested project structures. - Updated `listAllProjects` to utilize the new recursive function for improved project listing. - Enhanced `getMemoriesFromProject` and `findMemoryAcrossProjects` to support nested project structures. feat: Introduce spec context injection in hooks configuration - Added a new hook configuration for "Spec Context Injection" to load project specs based on prompt keywords. chore: Add gray-matter dependency for YAML frontmatter parsing - Included `gray-matter` package in `package.json` for parsing YAML frontmatter in markdown files. feat: Create Spec Index Builder tool for managing project specs - Implemented `spec-index-builder.ts` to scan markdown files, extract YAML frontmatter, and generate index cache files for different spec dimensions. feat: Develop Spec Init tool for initializing spec directories and seed documents - Created `spec-init.ts` to set up the directory structure and seed documents for the spec system. feat: Build Spec Keyword Extractor for keyword extraction from prompts - Added `spec-keyword-extractor.ts` to extract keywords from user prompts, supporting both English and Chinese text. feat: Implement Spec Loader for loading and filtering specs based on keywords - Developed `spec-loader.ts` to handle loading of specs, filtering by read mode and keyword matches, and formatting output for CLI or hooks.
16 KiB
16 KiB
name, description, allowed-tools
| name | description | allowed-tools |
|---|---|---|
| team-issue | Unified team skill for issue resolution. All roles invoke this skill with --role arg for role-specific execution. Triggers on "team issue". | TeamCreate(*), TeamDelete(*), SendMessage(*), TaskCreate(*), TaskUpdate(*), TaskList(*), TaskGet(*), Task(*), AskUserQuestion(*), Read(*), Write(*), Edit(*), Bash(*), Glob(*), Grep(*) |
Team Issue Resolution
Unified team skill for issue processing pipeline. All team members invoke this skill with --role=xxx for role-specific execution.
Scope: Issue 处理流程(plan → queue → execute)。Issue 创建/发现由 issue-discover 独立处理,CRUD 管理由 issue-manage 独立处理。
Architecture Overview
┌───────────────────────────────────────────┐
│ Skill(skill="team-issue") │
│ args="--role=xxx [issue-ids] [--mode=M]" │
└───────────────┬───────────────────────────┘
│ Role Router
┌───────────┼───────────┬───────────┬───────────┬───────────┐
↓ ↓ ↓ ↓ ↓ ↓
┌──────────┐┌──────────┐┌──────────┐┌──────────┐┌──────────┐┌──────────┐
│coordinator││explorer ││planner ││reviewer ││integrator││implementer│
│ 编排调度 ││EXPLORE-* ││SOLVE-* ││AUDIT-* ││MARSHAL-* ││BUILD-* │
└──────────┘└──────────┘└──────────┘└──────────┘└──────────┘└──────────┘
Command Architecture
roles/
├── coordinator.md # Pipeline 编排(Phase 1/5 inline, Phase 2-4 core logic)
├── explorer.md # 上下文分析(ACE + cli-explore-agent)
├── planner.md # 方案设计(wraps issue-plan-agent)
├── reviewer.md # 方案审查(技术可行性 + 风险评估)
├── integrator.md # 队列编排(wraps issue-queue-agent)
└── implementer.md # 代码实现(wraps code-developer)
Role Router
Input Parsing
Parse $ARGUMENTS to extract --role:
const args = "$ARGUMENTS"
const roleMatch = args.match(/--role[=\s]+(\w+)/)
if (!roleMatch) {
// No --role: this is a coordinator entry point
// Extract issue IDs and mode from args directly
// → Read roles/coordinator.md and execute
Read("roles/coordinator.md")
return
}
const role = roleMatch[1]
const teamName = "issue"
const agentName = args.match(/--agent-name[=\s]+([\w-]+)/)?.[1] || role
Role Dispatch
const VALID_ROLES = {
"coordinator": { file: "roles/coordinator.md", prefix: null },
"explorer": { file: "roles/explorer.md", prefix: "EXPLORE" },
"planner": { file: "roles/planner.md", prefix: "SOLVE" },
"reviewer": { file: "roles/reviewer.md", prefix: "AUDIT" },
"integrator": { file: "roles/integrator.md", prefix: "MARSHAL" },
"implementer": { file: "roles/implementer.md", prefix: "BUILD" }
}
if (!VALID_ROLES[role]) {
throw new Error(`Unknown role: ${role}. Available: ${Object.keys(VALID_ROLES).join(', ')}`)
}
// Read and execute role-specific logic
Read(VALID_ROLES[role].file)
// → Execute the 5-phase process defined in that file
Available Roles
| Role | Task Prefix | Responsibility | Reuses Agent | Role File |
|---|---|---|---|---|
coordinator |
N/A | Pipeline 编排、模式选择、任务分派 | - | roles/coordinator.md |
explorer |
EXPLORE-* | 上下文分析、影响面评估 | cli-explore-agent | roles/explorer.md |
planner |
SOLVE-* | 方案设计、任务分解 | issue-plan-agent | roles/planner.md |
reviewer |
AUDIT-* | 方案审查、风险评估 | - (新角色) | roles/reviewer.md |
integrator |
MARSHAL-* | 冲突检测、队列编排 | issue-queue-agent | roles/integrator.md |
implementer |
BUILD-* | 代码实现、结果提交 | code-developer | roles/implementer.md |
Shared Infrastructure
Role Isolation Rules
核心原则: 每个角色仅能执行自己职责范围内的工作。
Output Tagging(强制)
所有角色的输出必须带 [role_name] 标识前缀:
// SendMessage — content 和 summary 都必须带标识
SendMessage({
content: `## [${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 issue-plan-agent 等实现类 agent |
| 分发任务给 worker | ❌ 直接执行分析/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) |
| 委派给复用的 agent | ❌ 修改不属于本职责的资源 |
Team Configuration
const TEAM_CONFIG = {
name: "issue",
sessionDir: ".workflow/.team-plan/issue/",
issueDataDir: ".workflow/issues/"
}
Message Bus (All Roles)
Every SendMessage before, must call mcp__ccw-tools__team_msg to log:
mcp__ccw-tools__team_msg({
operation: "log",
team: "issue",
from: role, // current role name
to: "coordinator",
type: "<type>",
summary: "[role] <summary>",
ref: "<file_path>" // optional
})
Message types by role:
| Role | Types |
|---|---|
| coordinator | task_assigned, pipeline_update, escalation, shutdown, error |
| explorer | context_ready, impact_assessed, error |
| planner | solution_ready, multi_solution, error |
| reviewer | approved, rejected, concerns, error |
| integrator | queue_ready, conflict_found, error |
| implementer | impl_complete, impl_failed, error |
CLI 回退
当 mcp__ccw-tools__team_msg MCP 不可用时:
Bash(`ccw team log --team "issue" --from "${role}" --to "coordinator" --type "<type>" --summary "<摘要>" --json`)
Task Lifecycle (All Roles)
// Standard task lifecycle every role follows
// Phase 1: Discovery
const tasks = TaskList()
const myTasks = tasks.filter(t =>
t.subject.startsWith(`${VALID_ROLES[role].prefix}-`) &&
t.owner === agentName && // Use agentName (e.g., 'explorer-1') instead of 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 2-4: Role-specific (see roles/{role}.md)
// Phase 5: Report + Loop — 所有输出必须带 [role] 标识
mcp__ccw-tools__team_msg({ operation: "log", team: "issue", 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
Pipeline Modes
Quick Mode (1-2 simple issues):
EXPLORE-001 → SOLVE-001 → MARSHAL-001 → BUILD-001
Full Mode (complex issues, with review):
EXPLORE-001 → SOLVE-001 → AUDIT-001 ─┬─(approved)→ MARSHAL-001 → BUILD-001..N(parallel)
└─(rejected)→ SOLVE-fix → AUDIT-002(re-review, max 2x)
Batch Mode (5-100 issues):
EXPLORE-001..N(batch≤5) → SOLVE-001..N(batch≤3) → AUDIT-001(batch) → MARSHAL-001 → BUILD-001..M(DAG parallel)
Mode Auto-Detection
function detectMode(issueIds, userMode) {
if (userMode) return userMode // 用户显式指定
const count = issueIds.length
if (count <= 2) {
// Check complexity via issue priority
const issues = issueIds.map(id => JSON.parse(Bash(`ccw issue status ${id} --json`)))
const hasHighPriority = issues.some(i => i.priority >= 4)
return hasHighPriority ? 'full' : 'quick'
}
return 'batch'
}
Coordinator Spawn Template
TeamCreate({ team_name: "issue" })
// Explorer — conditional parallel spawn for Batch mode
const isBatchMode = mode === 'batch'
const maxParallelExplorers = Math.min(issueIds.length, 5)
if (isBatchMode && issueIds.length > 1) {
// Batch mode: spawn N explorers for parallel exploration
for (let i = 0; i < maxParallelExplorers; i++) {
const agentName = `explorer-${i + 1}`
Task({
subagent_type: "general-purpose",
description: `Spawn issue worker`,
team_name: "issue",
name: agentName,
prompt: `你是 team "issue" 的 EXPLORER (${agentName})。
你的 agent 名称是 "${agentName}",任务发现时用此名称匹配 owner。
当你收到 EXPLORE-* 任务时,调用 Skill(skill="team-issue", args="--role=explorer --agent-name=${agentName}") 执行。
当前需求: ${taskDescription}
约束: ${constraints}
## 角色准则(强制)
- 你只能处理 owner 为 "${agentName}" 的 EXPLORE-* 前缀任务
- 所有输出(SendMessage、team_msg)必须带 [explorer] 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:
1. TaskList → 找到 owner === "${agentName}" 的 EXPLORE-* 任务
2. Skill(skill="team-issue", args="--role=explorer --agent-name=${agentName}") 执行
3. team_msg log + SendMessage 结果给 coordinator(带 [explorer] 标识)
4. TaskUpdate completed → 检查下一个任务`
})
}
} else {
// Quick/Full mode: single explorer
Task({
subagent_type: "general-purpose",
description: `Spawn issue worker`,
team_name: "issue",
name: "explorer",
prompt: `你是 team "issue" 的 EXPLORER。
当你收到 EXPLORE-* 任务时,调用 Skill(skill="team-issue", args="--role=explorer") 执行。
当前需求: ${taskDescription}
约束: ${constraints}
## 角色准则(强制)
- 你只能处理 EXPLORE-* 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 [explorer] 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:
1. TaskList → 找到 EXPLORE-* 任务
2. Skill(skill="team-issue", args="--role=explorer") 执行
3. team_msg log + SendMessage 结果给 coordinator(带 [explorer] 标识)
4. TaskUpdate completed → 检查下一个任务`
})
}
// Planner
Task({
subagent_type: "general-purpose",
description: `Spawn issue worker`,
team_name: "issue",
name: "planner",
prompt: `你是 team "issue" 的 PLANNER。
当你收到 SOLVE-* 任务时,调用 Skill(skill="team-issue", args="--role=planner") 执行。
当前需求: ${taskDescription}
约束: ${constraints}
## 角色准则(强制)
- 你只能处理 SOLVE-* 前缀的任务
- 所有输出必须带 [planner] 标识前缀
- 仅与 coordinator 通信
## 消息总线(必须)
每次 SendMessage 前,先调用 mcp__ccw-tools__team_msg 记录。
工作流程:
1. TaskList → 找到 SOLVE-* 任务
2. Skill(skill="team-issue", args="--role=planner") 执行
3. team_msg log + SendMessage 结果给 coordinator
4. TaskUpdate completed → 检查下一个任务`
})
// Reviewer
Task({
subagent_type: "general-purpose",
description: `Spawn issue worker`,
team_name: "issue",
name: "reviewer",
prompt: `你是 team "issue" 的 REVIEWER。
当你收到 AUDIT-* 任务时,调用 Skill(skill="team-issue", args="--role=reviewer") 执行。
当前需求: ${taskDescription}
约束: ${constraints}
## 角色准则(强制)
- 你只能处理 AUDIT-* 前缀的任务
- 所有输出必须带 [reviewer] 标识前缀
- 仅与 coordinator 通信
- 你是质量门控角色,审查方案但不修改代码
## 消息总线(必须)
每次 SendMessage 前,先调用 mcp__ccw-tools__team_msg 记录。
工作流程:
1. TaskList → 找到 AUDIT-* 任务
2. Skill(skill="team-issue", args="--role=reviewer") 执行
3. team_msg log + SendMessage 结果给 coordinator
4. TaskUpdate completed → 检查下一个任务`
})
// Integrator
Task({
subagent_type: "general-purpose",
description: `Spawn issue worker`,
team_name: "issue",
name: "integrator",
prompt: `你是 team "issue" 的 INTEGRATOR。
当你收到 MARSHAL-* 任务时,调用 Skill(skill="team-issue", args="--role=integrator") 执行。
当前需求: ${taskDescription}
约束: ${constraints}
## 角色准则(强制)
- 你只能处理 MARSHAL-* 前缀的任务
- 所有输出必须带 [integrator] 标识前缀
- 仅与 coordinator 通信
## 消息总线(必须)
每次 SendMessage 前,先调用 mcp__ccw-tools__team_msg 记录。
工作流程:
1. TaskList → 找到 MARSHAL-* 任务
2. Skill(skill="team-issue", args="--role=integrator") 执行
3. team_msg log + SendMessage 结果给 coordinator
4. TaskUpdate completed → 检查下一个任务`
})
// Implementer — conditional parallel spawn for Batch mode (DAG parallel BUILD tasks)
if (isBatchMode && issueIds.length > 2) {
// Batch mode: spawn multiple implementers for parallel BUILD execution
const maxParallelBuilders = Math.min(issueIds.length, 3)
for (let i = 0; i < maxParallelBuilders; i++) {
const agentName = `implementer-${i + 1}`
Task({
subagent_type: "general-purpose",
description: `Spawn issue worker`,
team_name: "issue",
name: agentName,
prompt: `你是 team "issue" 的 IMPLEMENTER (${agentName})。
你的 agent 名称是 "${agentName}",任务发现时用此名称匹配 owner。
当你收到 BUILD-* 任务时,调用 Skill(skill="team-issue", args="--role=implementer --agent-name=${agentName}") 执行。
当前需求: ${taskDescription}
约束: ${constraints}
## 角色准则(强制)
- 你只能处理 owner 为 "${agentName}" 的 BUILD-* 前缀任务
- 所有输出必须带 [implementer] 标识前缀
- 仅与 coordinator 通信
## 消息总线(必须)
每次 SendMessage 前,先调用 mcp__ccw-tools__team_msg 记录。
工作流程:
1. TaskList → 找到 owner === "${agentName}" 的 BUILD-* 任务
2. Skill(skill="team-issue", args="--role=implementer --agent-name=${agentName}") 执行
3. team_msg log + SendMessage 结果给 coordinator
4. TaskUpdate completed → 检查下一个任务`
})
}
} else {
// Quick/Full mode: single implementer
Task({
subagent_type: "general-purpose",
description: `Spawn issue worker`,
team_name: "issue",
name: "implementer",
prompt: `你是 team "issue" 的 IMPLEMENTER。
当你收到 BUILD-* 任务时,调用 Skill(skill="team-issue", args="--role=implementer") 执行。
当前需求: ${taskDescription}
约束: ${constraints}
## 角色准则(强制)
- 你只能处理 BUILD-* 前缀的任务
- 所有输出必须带 [implementer] 标识前缀
- 仅与 coordinator 通信
## 消息总线(必须)
每次 SendMessage 前,先调用 mcp__ccw-tools__team_msg 记录。
工作流程:
1. TaskList → 找到 BUILD-* 任务
2. Skill(skill="team-issue", args="--role=implementer") 执行
3. team_msg log + SendMessage 结果给 coordinator
4. TaskUpdate completed → 检查下一个任务`
})
}
Error Handling
| Scenario | Resolution |
|---|---|
| Unknown --role value | Error with available role list |
| Missing --role arg | Default to coordinator role |
| Role file not found | Error with expected path (roles/{name}.md) |
| Task prefix conflict | Log warning, proceed |