mirror of
https://github.com/catlog22/Claude-Code-Workflow.git
synced 2026-02-14 02:42:04 +08:00
Add unified command indices for CCW and CCW-Help with detailed capabilities, flows, and intent rules
- Introduced command.json for CCW-Help with 88 commands and 16 agents, covering essential workflows and memory management. - Created command.json for CCW with comprehensive capabilities for exploration, planning, execution, bug fixing, testing, reviewing, and documentation. - Defined complex flows for rapid iteration, full exploration, coupled planning, bug fixing, issue lifecycle management, and more. - Implemented intent rules for bug fixing, issue batch processing, exploration, UI design, TDD, review, and documentation. - Established CLI tools and injection rules to enhance command execution based on context and complexity.
This commit is contained in:
catlog22
@@ -1,469 +1,352 @@
|
||||
---
|
||||
name: ccw
|
||||
description: Stateless workflow orchestrator that automatically selects and executes the optimal workflow combination based on task intent. Supports rapid (lite-plan+execute), full (brainstorm+plan+execute), coupled (plan+execute), bugfix (lite-fix), and issue (multi-point fixes) workflows. Triggers on "ccw", "workflow", "自动工作流", "智能调度".
|
||||
allowed-tools: Task(*), SlashCommand(*), AskUserQuestion(*), Read(*), Bash(*), Grep(*)
|
||||
description: Stateless workflow orchestrator. Auto-selects optimal workflow based on task intent. Triggers "ccw", "workflow".
|
||||
allowed-tools: Task(*), SlashCommand(*), AskUserQuestion(*), Read(*), Bash(*), Grep(*), TodoWrite(*)
|
||||
---
|
||||
|
||||
# CCW - Claude Code Workflow Orchestrator
|
||||
|
||||
无状态工作流协调器，根据任务意图自动选择并执行最优工作流组合。
|
||||
无状态工作流协调器，根据任务意图自动选择最优工作流。
|
||||
|
||||
## Architecture Overview
|
||||
## Architecture
|
||||
|
||||
```
|
||||
┌─────────────────────────────────────────────────────────────────┐
|
||||
│ CCW Orchestrator (Stateless + Requirement Analysis) │
|
||||
│ CCW Orchestrator (CLI-Enhanced + Requirement Analysis) │
|
||||
├─────────────────────────────────────────────────────────────────┤
|
||||
│ │
|
||||
│ Input Analysis & Requirement Extraction │
|
||||
│ ├─ Intent Classification (bugfix/feature/refactor/issue/...) │
|
||||
│ ├─ Complexity Assessment (low/medium/high) │
|
||||
│ ├─ Dimension Extraction (WHAT/WHERE/WHY/HOW) │
|
||||
│ ├─ Clarity Scoring (0-3) with auto-clarification │
|
||||
│ └─ Constraint Extraction (time/scope/quality) │
|
||||
│ │
|
||||
│ Workflow Selection (Decision Tree) │
|
||||
│ ├─ 🐛 Bug? → lite-fix / lite-fix --hotfix │
|
||||
│ ├─ ❓ Unclear? → clarify → brainstorm → plan → execute │
|
||||
│ ├─ ⚡ Simple? → lite-plan → lite-execute │
|
||||
│ ├─ 🔧 Complex? → plan → execute │
|
||||
│ ├─ 📋 Issue? → discover → plan → queue → execute │
|
||||
│ └─ 🎨 UI? → ui-design → plan → execute │
|
||||
│ │
|
||||
│ Execution Dispatch │
|
||||
│ └─ SlashCommand("/workflow:xxx") or Task(agent) │
|
||||
│ │
|
||||
│ Phase 1 │ Input Analysis (rule-based, fast path) │
|
||||
│ Phase 1.5 │ CLI Classification (semantic, smart path) │
|
||||
│ Phase 1.75 │ Requirement Clarification (clarity < 2) │
|
||||
│ Phase 2 │ Chain Selection (intent → workflow) │
|
||||
│ Phase 2.5 │ CLI Action Planning (high complexity) │
|
||||
│ Phase 3 │ User Confirmation (optional) │
|
||||
│ Phase 4 │ TODO Tracking Setup │
|
||||
│ Phase 5 │ Execution Loop │
|
||||
└─────────────────────────────────────────────────────────────────┘
|
||||
```
|
||||
|
||||
## Workflow Combinations (组合技)
|
||||
|
||||
### 1. Rapid (快速迭代) ⚡
|
||||
**Pattern**: 多模型协作分析 + 直接执行
|
||||
**Commands**: `/workflow:lite-plan` → `/workflow:lite-execute`
|
||||
**When to use**:
|
||||
- 明确知道做什么和怎么做
|
||||
- 单一功能或小型改动
|
||||
- 快速原型验证
|
||||
|
||||
### 2. Full (完整流程) 📋
|
||||
**Pattern**: 分析 + 头脑风暴 + 规划 + 执行
|
||||
**Commands**: `/workflow:brainstorm:auto-parallel` → `/workflow:plan` → `/workflow:execute`
|
||||
**When to use**:
|
||||
- 不确定产品方向或技术方案
|
||||
- 需要多角色视角分析
|
||||
- 复杂新功能开发
|
||||
|
||||
### 3. Coupled (复杂耦合) 🔗
|
||||
**Pattern**: 完整规划 + 验证 + 执行
|
||||
**Commands**: `/workflow:plan` → `/workflow:action-plan-verify` → `/workflow:execute`
|
||||
**When to use**:
|
||||
- 跨模块依赖
|
||||
- 架构级变更
|
||||
- 团队协作项目
|
||||
|
||||
### 4. Bugfix (缺陷修复) 🐛
|
||||
**Pattern**: 智能诊断 + 修复
|
||||
**Commands**: `/workflow:lite-fix` or `/workflow:lite-fix --hotfix`
|
||||
**When to use**:
|
||||
- 任何有明确症状的Bug
|
||||
- 生产事故紧急修复
|
||||
- 根因不清楚需要诊断
|
||||
|
||||
### 5. Issue (长时间多点修复) 📌
|
||||
**Pattern**: 发现 + 创建 + 规划 + 队列 + 批量执行
|
||||
**Commands**:
|
||||
- 完整链: `/issue:discover` → `/issue:new` → `/issue:plan` → `/issue:queue` → `/issue:execute`
|
||||
- 快速链: `/issue:plan` → `/issue:queue` → `/issue:execute`
|
||||
- 提示发现: `/issue:discover-by-prompt` → `/issue:plan` → `/issue:queue` → `/issue:execute`
|
||||
**When to use**:
|
||||
- 多个相关问题需要批量处理
|
||||
- 长时间跨度的修复任务
|
||||
- 需要优先级排序和冲突解决
|
||||
- 安全审计、技术债务清理、GitHub Issues 批量导入
|
||||
|
||||
### 6. UI-First (设计驱动) 🎨
|
||||
**Pattern**: UI设计 + 规划 + 执行
|
||||
**Commands**: `/workflow:ui-design:*` → `/workflow:plan` → `/workflow:execute`
|
||||
**When to use**:
|
||||
- 前端功能开发
|
||||
- 需要视觉参考
|
||||
- 设计系统集成
|
||||
|
||||
## Intent Classification
|
||||
|
||||
```javascript
|
||||
function classifyIntent(input) {
|
||||
const text = input.toLowerCase()
|
||||
|
||||
// Priority 1: Bug keywords
|
||||
if (/\b(fix|bug|error|issue|crash|broken|fail|wrong|incorrect)\b/.test(text)) {
|
||||
if (/\b(hotfix|urgent|production|critical|emergency)\b/.test(text)) {
|
||||
return { type: 'bugfix', mode: 'hotfix', workflow: 'lite-fix --hotfix' }
|
||||
}
|
||||
return { type: 'bugfix', mode: 'standard', workflow: 'lite-fix' }
|
||||
}
|
||||
|
||||
// Priority 2: Issue batch keywords
|
||||
if (/\b(issues?|batch|queue|多个|批量)\b/.test(text) && /\b(fix|resolve|处理)\b/.test(text)) {
|
||||
return { type: 'issue', workflow: 'issue:plan → issue:queue → issue:execute' }
|
||||
}
|
||||
|
||||
// Priority 3: Uncertainty keywords → Full workflow
|
||||
if (/\b(不确定|不知道|explore|研究|分析一下|怎么做|what if|should i|探索)\b/.test(text)) {
|
||||
return { type: 'exploration', workflow: 'brainstorm → plan → execute' }
|
||||
}
|
||||
|
||||
// Priority 4: UI/Design keywords
|
||||
if (/\b(ui|界面|design|设计|component|组件|style|样式|layout|布局)\b/.test(text)) {
|
||||
return { type: 'ui', workflow: 'ui-design → plan → execute' }
|
||||
}
|
||||
|
||||
// Priority 5: Complexity assessment for remaining
|
||||
const complexity = assessComplexity(text)
|
||||
|
||||
if (complexity === 'high') {
|
||||
return { type: 'feature', complexity: 'high', workflow: 'plan → verify → execute' }
|
||||
}
|
||||
|
||||
if (complexity === 'medium') {
|
||||
return { type: 'feature', complexity: 'medium', workflow: 'lite-plan → lite-execute' }
|
||||
}
|
||||
|
||||
return { type: 'feature', complexity: 'low', workflow: 'lite-plan → lite-execute' }
|
||||
}
|
||||
### Priority Order
|
||||
|
||||
| Priority | Intent | Patterns | Flow |
|
||||
|----------|--------|----------|------|
|
||||
| 1 | bugfix/hotfix | `urgent,production,critical` + bug | `bugfix.hotfix` |
|
||||
| 1 | bugfix | `fix,bug,error,crash,fail` | `bugfix.standard` |
|
||||
| 2 | issue batch | `issues,batch` + `fix,resolve` | `issue` |
|
||||
| 3 | exploration | `不确定,explore,研究,what if` | `full` |
|
||||
| 3 | multi-perspective | `多视角,权衡,比较方案,cross-verify` | `multi-cli-plan` |
|
||||
| 4 | quick-task | `快速,简单,small,quick` + feature | `lite-lite-lite` |
|
||||
| 5 | ui design | `ui,design,component,style` | `ui` |
|
||||
| 6 | tdd | `tdd,test-driven,先写测试` | `tdd` |
|
||||
| 7 | review | `review,审查,code review` | `review-fix` |
|
||||
| 8 | documentation | `文档,docs,readme` | `docs` |
|
||||
| 99 | feature | complexity-based | `rapid`/`coupled` |
|
||||
|
||||
### Complexity Assessment
|
||||
|
||||
```javascript
|
||||
function assessComplexity(text) {
|
||||
let score = 0
|
||||
|
||||
// Architecture keywords
|
||||
if (/\b(refactor|重构|migrate|迁移|architect|架构|system|系统)\b/.test(text)) score += 2
|
||||
|
||||
// Multi-module keywords
|
||||
if (/\b(multiple|多个|across|跨|all|所有|entire|整个)\b/.test(text)) score += 2
|
||||
|
||||
// Integration keywords
|
||||
if (/\b(integrate|集成|connect|连接|api|database|数据库)\b/.test(text)) score += 1
|
||||
|
||||
// Security/Performance keywords
|
||||
if (/\b(security|安全|performance|性能|scale|扩展)\b/.test(text)) score += 1
|
||||
|
||||
if (score >= 4) return 'high'
|
||||
if (score >= 2) return 'medium'
|
||||
return 'low'
|
||||
if (/refactor|重构|migrate|迁移|architect|架构|system|系统/.test(text)) score += 2
|
||||
if (/multiple|多个|across|跨|all|所有|entire|整个/.test(text)) score += 2
|
||||
if (/integrate|集成|api|database|数据库/.test(text)) score += 1
|
||||
if (/security|安全|performance|性能|scale|扩展/.test(text)) score += 1
|
||||
return score >= 4 ? 'high' : score >= 2 ? 'medium' : 'low'
|
||||
}
|
||||
```
|
||||
|
||||
## Execution Flow
|
||||
| Complexity | Flow |
|
||||
|------------|------|
|
||||
| high | `coupled` (plan → verify → execute) |
|
||||
| medium/low | `rapid` (lite-plan → lite-execute) |
|
||||
|
||||
### Phase 1: Input Analysis
|
||||
### Dimension Extraction (WHAT/WHERE/WHY/HOW)
|
||||
|
||||
从用户输入提取四个维度，用于需求澄清和工作流选择：
|
||||
|
||||
| 维度 | 提取内容 | 示例模式 |
|
||||
|------|----------|----------|
|
||||
| **WHAT** | action + target | `创建/修复/重构/优化/分析` + 目标对象 |
|
||||
| **WHERE** | scope + paths | `file/module/system` + 文件路径 |
|
||||
| **WHY** | goal + motivation | `为了.../因为.../目的是...` |
|
||||
| **HOW** | constraints + preferences | `必须.../不要.../应该...` |
|
||||
|
||||
**Clarity Score** (0-3):
|
||||
- +0.5: 有明确 action
|
||||
- +0.5: 有具体 target
|
||||
- +0.5: 有文件路径
|
||||
- +0.5: scope 不是 unknown
|
||||
- +0.5: 有明确 goal
|
||||
- +0.5: 有约束条件
|
||||
- -0.5: 包含不确定词 (`不知道/maybe/怎么`)
|
||||
|
||||
### Requirement Clarification
|
||||
|
||||
当 `clarity_score < 2` 时触发需求澄清：
|
||||
|
||||
```javascript
|
||||
// Parse user input
|
||||
const input = userInput.trim()
|
||||
if (dimensions.clarity_score < 2) {
|
||||
const questions = generateClarificationQuestions(dimensions)
|
||||
// 生成问题：目标是什么? 范围是什么? 有什么约束?
|
||||
AskUserQuestion({ questions })
|
||||
}
|
||||
```
|
||||
|
||||
// Check for explicit workflow request
|
||||
**澄清问题类型**:
|
||||
- 目标不明确 → "你想要对什么进行操作?"
|
||||
- 范围不明确 → "操作的范围是什么?"
|
||||
- 目的不明确 → "这个操作的主要目标是什么?"
|
||||
- 复杂操作 → "有什么特殊要求或限制?"
|
||||
|
||||
## TODO Tracking Protocol
|
||||
|
||||
### CRITICAL: Append-Only Rule
|
||||
|
||||
CCW 创建的 Todo **必须附加到现有列表**，不能覆盖用户的其他 Todo。
|
||||
|
||||
### Implementation
|
||||
|
||||
```javascript
|
||||
// 1. 使用 CCW 前缀隔离工作流 todo
|
||||
const prefix = `CCW:${flowName}`
|
||||
|
||||
// 2. 创建新 todo 时使用前缀格式
|
||||
TodoWrite({
|
||||
todos: [
|
||||
...existingNonCCWTodos, // 保留用户的 todo
|
||||
{ content: `${prefix}: [1/N] /command:step1`, status: "in_progress", activeForm: "..." },
|
||||
{ content: `${prefix}: [2/N] /command:step2`, status: "pending", activeForm: "..." }
|
||||
]
|
||||
})
|
||||
|
||||
// 3. 更新状态时只修改匹配前缀的 todo
|
||||
```
|
||||
|
||||
### Todo Format
|
||||
|
||||
```
|
||||
CCW:{flow}: [{N}/{Total}] /command:name
|
||||
```
|
||||
|
||||
### Visual Example
|
||||
|
||||
```
|
||||
✓ CCW:rapid: [1/2] /workflow:lite-plan
|
||||
→ CCW:rapid: [2/2] /workflow:lite-execute
|
||||
用户自己的 todo（保留不动）
|
||||
```
|
||||
|
||||
### Status Management
|
||||
|
||||
- 开始工作流：创建所有步骤 todo，第一步 `in_progress`
|
||||
- 完成步骤：当前步骤 `completed`，下一步 `in_progress`
|
||||
- 工作流结束：所有 CCW todo 标记 `completed`
|
||||
|
||||
## Execution Flow
|
||||
|
||||
```javascript
|
||||
// 1. Check explicit command
|
||||
if (input.startsWith('/workflow:') || input.startsWith('/issue:')) {
|
||||
// User explicitly requested a workflow, pass through
|
||||
SlashCommand(input)
|
||||
return
|
||||
}
|
||||
|
||||
// Classify intent
|
||||
const intent = classifyIntent(input)
|
||||
// 2. Classify intent
|
||||
const intent = classifyIntent(input) // See command.json intent_rules
|
||||
|
||||
console.log(`
|
||||
## Intent Analysis
|
||||
// 3. Select flow
|
||||
const flow = selectFlow(intent) // See command.json flows
|
||||
|
||||
**Input**: ${input.substring(0, 100)}...
|
||||
**Classification**: ${intent.type}
|
||||
**Complexity**: ${intent.complexity || 'N/A'}
|
||||
**Recommended Workflow**: ${intent.workflow}
|
||||
`)
|
||||
```
|
||||
// 4. Create todos with CCW prefix
|
||||
createWorkflowTodos(flow)
|
||||
|
||||
### Phase 2: User Confirmation (Optional)
|
||||
|
||||
```javascript
|
||||
// For high-complexity or ambiguous intents, confirm with user
|
||||
if (intent.complexity === 'high' || intent.type === 'exploration') {
|
||||
const confirmation = AskUserQuestion({
|
||||
questions: [{
|
||||
question: `Recommended: ${intent.workflow}. Proceed?`,
|
||||
header: "Workflow",
|
||||
multiSelect: false,
|
||||
options: [
|
||||
{ label: `${intent.workflow} (Recommended)`, description: "Use recommended workflow" },
|
||||
{ label: "Rapid (lite-plan)", description: "Quick iteration" },
|
||||
{ label: "Full (brainstorm+plan)", description: "Complete exploration" },
|
||||
{ label: "Manual", description: "I'll specify the commands" }
|
||||
]
|
||||
}]
|
||||
})
|
||||
|
||||
// Adjust workflow based on user selection
|
||||
intent.workflow = mapSelectionToWorkflow(confirmation)
|
||||
}
|
||||
```
|
||||
|
||||
### Phase 3: Workflow Dispatch
|
||||
|
||||
```javascript
|
||||
switch (intent.workflow) {
|
||||
case 'lite-fix':
|
||||
SlashCommand('/workflow:lite-fix', args: input)
|
||||
break
|
||||
|
||||
case 'lite-fix --hotfix':
|
||||
SlashCommand('/workflow:lite-fix --hotfix', args: input)
|
||||
break
|
||||
|
||||
case 'lite-plan → lite-execute':
|
||||
SlashCommand('/workflow:lite-plan', args: input)
|
||||
// lite-plan will automatically dispatch to lite-execute
|
||||
break
|
||||
|
||||
case 'plan → verify → execute':
|
||||
SlashCommand('/workflow:plan', args: input)
|
||||
// After plan, prompt for verify and execute
|
||||
break
|
||||
|
||||
case 'brainstorm → plan → execute':
|
||||
SlashCommand('/workflow:brainstorm:auto-parallel', args: input)
|
||||
// After brainstorm, continue with plan
|
||||
break
|
||||
|
||||
case 'issue:plan → issue:queue → issue:execute':
|
||||
SlashCommand('/issue:plan', args: input)
|
||||
// Issue workflow handles queue and execute
|
||||
break
|
||||
|
||||
case 'ui-design → plan → execute':
|
||||
// Determine UI design subcommand
|
||||
if (hasReference(input)) {
|
||||
SlashCommand('/workflow:ui-design:imitate-auto', args: input)
|
||||
} else {
|
||||
SlashCommand('/workflow:ui-design:explore-auto', args: input)
|
||||
}
|
||||
break
|
||||
}
|
||||
// 5. Dispatch first command
|
||||
SlashCommand(flow.steps[0].command, args: input)
|
||||
```
|
||||
|
||||
## CLI Tool Integration
|
||||
|
||||
CCW **隐式调用** CLI 工具以获得三大优势：
|
||||
CCW 在特定条件下自动注入 CLI 调用：
|
||||
|
||||
### 1. Token 效率 (Context Efficiency)
|
||||
| Condition | CLI Inject |
|
||||
|-----------|------------|
|
||||
| 大量代码上下文 (≥50k chars) | `gemini --mode analysis` |
|
||||
| 高复杂度任务 | `gemini --mode analysis` |
|
||||
| Bug 诊断 | `gemini --mode analysis` |
|
||||
| 多任务执行 (≥3 tasks) | `codex --mode write` |
|
||||
|
||||
CLI 工具在单独进程中运行，可以处理大量代码上下文而不消耗主会话 token：
|
||||
### CLI Enhancement Phases
|
||||
|
||||
| 场景 | 触发条件 | 自动注入 |
|
||||
|------|----------|----------|
|
||||
| 大量代码上下文 | 文件读取 ≥ 50k 字符 | `gemini --mode analysis` |
|
||||
| 多模块分析 | 涉及 ≥ 5 个模块 | `gemini --mode analysis` |
|
||||
| 代码审查 | review 步骤 | `gemini --mode analysis` |
|
||||
**Phase 1.5: CLI-Assisted Classification**
|
||||
|
||||
### 2. 多模型视角 (Multi-Model Perspectives)
|
||||
当规则匹配不明确时，使用 CLI 辅助分类：
|
||||
|
||||
不同模型有不同优势，CCW 根据任务类型自动选择：
|
||||
| 触发条件 | 说明 |
|
||||
|----------|------|
|
||||
| matchCount < 2 | 多个意图模式匹配 |
|
||||
| complexity = high | 高复杂度任务 |
|
||||
| input > 100 chars | 长输入需要语义理解 |
|
||||
|
||||
| Tool | 核心优势 | 最佳场景 | 触发关键词 |
|
||||
|------|----------|----------|------------|
|
||||
| Gemini | 超长上下文、深度分析、架构理解、执行流追踪 | 代码库理解、架构评估、根因分析 | "分析", "理解", "设计", "架构", "诊断" |
|
||||
| Qwen | 代码模式识别、多维度分析 | Gemini备选、第二视角验证 | "评估", "对比", "验证" |
|
||||
| Codex | 精确代码生成、自主执行、数学推理 | 功能实现、重构、测试 | "实现", "重构", "修复", "生成", "测试" |
|
||||
**Phase 2.5: CLI-Assisted Action Planning**
|
||||
|
||||
### 3. 增强能力 (Enhanced Capabilities)
|
||||
高复杂度任务的工作流优化：
|
||||
|
||||
#### Debug 能力增强
|
||||
```
|
||||
触发条件: intent === 'bugfix' AND root_cause_unclear
|
||||
自动注入: gemini --mode analysis (执行流追踪)
|
||||
用途: 假设驱动调试、状态机错误诊断、并发问题排查
|
||||
```
|
||||
| 触发条件 | 说明 |
|
||||
|----------|------|
|
||||
| complexity = high | 高复杂度任务 |
|
||||
| steps >= 3 | 多步骤工作流 |
|
||||
| input > 200 chars | 复杂需求描述 |
|
||||
|
||||
#### 规划能力增强
|
||||
```
|
||||
触发条件: complexity === 'high' OR intent === 'exploration'
|
||||
自动注入: gemini --mode analysis (架构分析)
|
||||
用途: 复杂任务先用CLI分析获取多模型视角
|
||||
```
|
||||
CLI 可返回建议：`use_default` | `modify` (调整步骤) | `upgrade` (升级工作流)
|
||||
|
||||
### 隐式注入规则 (Implicit Injection Rules)
|
||||
## Continuation Commands
|
||||
|
||||
CCW 在以下条件自动注入 CLI 调用（无需用户显式请求）：
|
||||
工作流执行中的用户控制命令：
|
||||
|
||||
```javascript
|
||||
const implicitRules = {
|
||||
// 上下文收集：大量代码使用CLI可节省主会话token
|
||||
context_gathering: {
|
||||
trigger: 'file_read >= 50k chars OR module_count >= 5',
|
||||
inject: 'gemini --mode analysis'
|
||||
},
|
||||
|
||||
// 规划前分析：复杂任务先用CLI分析
|
||||
pre_planning_analysis: {
|
||||
trigger: 'complexity === "high" OR intent === "exploration"',
|
||||
inject: 'gemini --mode analysis'
|
||||
},
|
||||
|
||||
// 调试诊断：利用Gemini的执行流追踪能力
|
||||
debug_diagnosis: {
|
||||
trigger: 'intent === "bugfix" AND root_cause_unclear',
|
||||
inject: 'gemini --mode analysis'
|
||||
},
|
||||
|
||||
// 代码审查：用CLI减少token占用
|
||||
code_review: {
|
||||
trigger: 'step === "review"',
|
||||
inject: 'gemini --mode analysis'
|
||||
},
|
||||
|
||||
// 多任务执行：用Codex自主完成
|
||||
implementation: {
|
||||
trigger: 'step === "execute" AND task_count >= 3',
|
||||
inject: 'codex --mode write'
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
### 用户语义触发 (Semantic Tool Assignment)
|
||||
|
||||
```javascript
|
||||
// 用户可以通过自然语言指定工具偏好
|
||||
const toolHints = {
|
||||
gemini: /用\s*gemini|gemini\s*分析|让\s*gemini|深度分析|架构理解/i,
|
||||
qwen: /用\s*qwen|qwen\s*评估|让\s*qwen|第二视角/i,
|
||||
codex: /用\s*codex|codex\s*实现|让\s*codex|自主完成|批量修改/i
|
||||
}
|
||||
|
||||
function detectToolPreference(input) {
|
||||
for (const [tool, pattern] of Object.entries(toolHints)) {
|
||||
if (pattern.test(input)) return tool
|
||||
}
|
||||
return null // Auto-select based on task type
|
||||
}
|
||||
```
|
||||
|
||||
### 独立 CLI 工作流 (Standalone CLI Workflows)
|
||||
|
||||
直接调用 CLI 进行特定任务：
|
||||
|
||||
| Workflow | 命令 | 用途 |
|
||||
|----------|------|------|
|
||||
| CLI Analysis | `ccw cli --tool gemini` | 大型代码库快速理解、架构评估 |
|
||||
| CLI Implement | `ccw cli --tool codex` | 明确需求的自主实现 |
|
||||
| CLI Debug | `ccw cli --tool gemini` | 复杂bug根因分析、执行流追踪 |
|
||||
|
||||
## Index Files (Dynamic Coordination)
|
||||
|
||||
CCW 使用索引文件实现智能命令协调：
|
||||
|
||||
| Index | Purpose |
|
||||
|-------|---------|
|
||||
| [index/command-capabilities.json](index/command-capabilities.json) | 命令能力分类（explore, plan, execute, test, review...） |
|
||||
| [index/workflow-chains.json](index/workflow-chains.json) | 预定义工作流链（rapid, full, coupled, bugfix, issue, tdd, ui...） |
|
||||
|
||||
### 能力分类
|
||||
|
||||
```
|
||||
capabilities:
|
||||
├── explore - 代码探索、上下文收集
|
||||
├── brainstorm - 多角色分析、方案探索
|
||||
├── plan - 任务规划、分解
|
||||
├── verify - 计划验证、质量检查
|
||||
├── execute - 任务执行、代码实现
|
||||
├── bugfix - Bug诊断、修复
|
||||
├── test - 测试生成、执行
|
||||
├── review - 代码审查、质量分析
|
||||
├── issue - 批量问题管理
|
||||
├── ui-design - UI设计、原型
|
||||
├── memory - 文档、知识管理
|
||||
├── session - 会话管理
|
||||
└── debug - 调试、问题排查
|
||||
```
|
||||
|
||||
## TODO Tracking Integration
|
||||
|
||||
CCW 自动使用 TodoWrite 跟踪工作流执行进度：
|
||||
|
||||
```javascript
|
||||
// 工作流启动时自动创建 TODO 列表
|
||||
TodoWrite({
|
||||
todos: [
|
||||
{ content: "CCW: Rapid Iteration (2 steps)", status: "in_progress", activeForm: "Running workflow" },
|
||||
{ content: "[1/2] /workflow:lite-plan", status: "in_progress", activeForm: "Executing lite-plan" },
|
||||
{ content: "[2/2] /workflow:lite-execute", status: "pending", activeForm: "Executing lite-execute" }
|
||||
]
|
||||
})
|
||||
|
||||
// 每个步骤完成后自动更新状态
|
||||
// 支持暂停、继续、跳过操作
|
||||
```
|
||||
|
||||
**进度可视化**:
|
||||
```
|
||||
✓ CCW: Rapid Iteration (2 steps)
|
||||
✓ [1/2] /workflow:lite-plan
|
||||
→ [2/2] /workflow:lite-execute
|
||||
```
|
||||
|
||||
**控制命令**:
|
||||
| Input | Action |
|
||||
|-------|--------|
|
||||
| `continue` | 执行下一步 |
|
||||
| 命令 | 作用 |
|
||||
|------|------|
|
||||
| `continue` | 继续执行下一步 |
|
||||
| `skip` | 跳过当前步骤 |
|
||||
| `abort` | 停止工作流 |
|
||||
| `/workflow:*` | 执行指定命令 |
|
||||
| `abort` | 终止工作流 |
|
||||
| `/workflow:*` | 切换到指定命令 |
|
||||
| 自然语言 | 重新分析意图 |
|
||||
|
||||
## Reference Documents
|
||||
## Workflow Flow Details
|
||||
|
||||
| Document | Purpose |
|
||||
|----------|---------|
|
||||
| [phases/orchestrator.md](phases/orchestrator.md) | 编排器决策逻辑 + 澄清流程 + TODO 跟踪 |
|
||||
| [phases/actions/rapid.md](phases/actions/rapid.md) | 快速迭代组合 |
|
||||
| [phases/actions/full.md](phases/actions/full.md) | 完整流程组合 |
|
||||
| [phases/actions/coupled.md](phases/actions/coupled.md) | 复杂耦合组合 |
|
||||
| [phases/actions/bugfix.md](phases/actions/bugfix.md) | 缺陷修复组合 |
|
||||
| [phases/actions/issue.md](phases/actions/issue.md) | Issue工作流组合 (完整链路) |
|
||||
| [specs/intent-classification.md](specs/intent-classification.md) | 意图分类规范 |
|
||||
| [specs/requirement-analysis.md](specs/requirement-analysis.md) | 需求分析规范 (NEW) |
|
||||
| [specs/output-templates.md](specs/output-templates.md) | 输出格式模板 (NEW) |
|
||||
| [WORKFLOW_DECISION_GUIDE.md](/WORKFLOW_DECISION_GUIDE.md) | 工作流决策指南 |
|
||||
### Issue Workflow (两阶段生命周期)
|
||||
|
||||
## Examples
|
||||
Issue 工作流设计为两阶段生命周期，支持在项目迭代过程中积累问题并集中解决。
|
||||
|
||||
**Phase 1: Accumulation (积累阶段)**
|
||||
- 触发：任务完成后的 review、代码审查发现、测试失败
|
||||
- 活动：需求扩展、bug 分析、测试覆盖、安全审查
|
||||
- 命令：`/issue:discover`, `/issue:discover-by-prompt`, `/issue:new`
|
||||
|
||||
**Phase 2: Batch Resolution (批量解决阶段)**
|
||||
- 触发：积累足够 issue 后的集中处理
|
||||
- 流程：plan → queue → execute
|
||||
- 命令：`/issue:plan --all-pending` → `/issue:queue` → `/issue:execute`
|
||||
|
||||
### Example 1: Bug Fix
|
||||
```
|
||||
User: 用户登录失败，返回 401 错误
|
||||
CCW: Intent=bugfix, Workflow=lite-fix
|
||||
→ /workflow:lite-fix "用户登录失败，返回 401 错误"
|
||||
任务完成 → discover → 积累 issue → ... → plan all → queue → parallel execute
|
||||
↑ ↓
|
||||
└────── 迭代循环 ───────┘
|
||||
```
|
||||
|
||||
### Example 2: New Feature (Simple)
|
||||
### lite-lite-lite vs multi-cli-plan
|
||||
|
||||
| 维度 | lite-lite-lite | multi-cli-plan |
|
||||
|------|---------------|----------------|
|
||||
| **产物** | 无文件 | IMPL_PLAN.md + plan.json + synthesis.json |
|
||||
| **状态** | 无状态 | 持久化 session |
|
||||
| **CLI选择** | 自动分析任务类型选择 | 配置驱动 |
|
||||
| **迭代** | 通过 AskUser | 多轮收敛 |
|
||||
| **执行** | 直接执行 | 通过 lite-execute |
|
||||
| **适用** | 快速修复、简单功能 | 复杂多步骤实现 |
|
||||
|
||||
**选择指南**：
|
||||
- 任务清晰、改动范围小 → `lite-lite-lite`
|
||||
- 需要多视角分析、复杂架构 → `multi-cli-plan`
|
||||
|
||||
### multi-cli-plan vs lite-plan
|
||||
|
||||
| 维度 | multi-cli-plan | lite-plan |
|
||||
|------|---------------|-----------|
|
||||
| **上下文** | ACE 语义搜索 | 手动文件模式 |
|
||||
| **分析** | 多 CLI 交叉验证 | 单次规划 |
|
||||
| **迭代** | 多轮直到收敛 | 单轮 |
|
||||
| **置信度** | 高 (共识驱动) | 中 (单一视角) |
|
||||
| **适用** | 需要多视角的复杂任务 | 直接明确的实现 |
|
||||
|
||||
**选择指南**：
|
||||
- 需求明确、路径清晰 → `lite-plan`
|
||||
- 需要权衡、多方案比较 → `multi-cli-plan`
|
||||
|
||||
## Artifact Flow Protocol
|
||||
|
||||
工作流产出的自动流转机制，支持不同格式产出间的意图提取和完成度判断。
|
||||
|
||||
### 产出格式
|
||||
|
||||
| 命令 | 产出位置 | 格式 | 关键字段 |
|
||||
|------|----------|------|----------|
|
||||
| `/workflow:lite-plan` | memory://plan | structured_plan | tasks, files, dependencies |
|
||||
| `/workflow:plan` | .workflow/{session}/IMPL_PLAN.md | markdown_plan | phases, tasks, risks |
|
||||
| `/workflow:execute` | execution_log.json | execution_report | completed_tasks, errors |
|
||||
| `/workflow:test-cycle-execute` | test_results.json | test_report | pass_rate, failures, coverage |
|
||||
| `/workflow:review-session-cycle` | review_report.md | review_report | findings, severity_counts |
|
||||
|
||||
### 意图提取 (Intent Extraction)
|
||||
|
||||
流转到下一步时，自动提取关键信息：
|
||||
|
||||
```
|
||||
User: 添加用户头像上传功能
|
||||
CCW: Intent=feature, Complexity=low, Workflow=lite-plan→lite-execute
|
||||
→ /workflow:lite-plan "添加用户头像上传功能"
|
||||
plan → execute:
|
||||
提取: tasks (未完成), priority_order, files_to_modify, context_summary
|
||||
|
||||
execute → test:
|
||||
提取: modified_files, test_scope (推断), pending_verification
|
||||
|
||||
test → fix:
|
||||
条件: pass_rate < 0.95
|
||||
提取: failures, error_messages, affected_files, suggested_fixes
|
||||
|
||||
review → fix:
|
||||
条件: critical > 0 OR high > 3
|
||||
提取: findings (critical/high), fix_priority, affected_files
|
||||
```
|
||||
|
||||
### Example 3: Complex Refactoring
|
||||
### 完成度判断
|
||||
|
||||
**Test 完成度路由**:
|
||||
```
|
||||
User: 重构整个认证模块，迁移到 OAuth2
|
||||
CCW: Intent=feature, Complexity=high, Workflow=plan→verify→execute
|
||||
→ /workflow:plan "重构整个认证模块，迁移到 OAuth2"
|
||||
pass_rate >= 0.95 AND coverage >= 0.80 → complete
|
||||
pass_rate >= 0.95 AND coverage < 0.80 → add_more_tests
|
||||
pass_rate >= 0.80 → fix_failures_then_continue
|
||||
pass_rate < 0.80 → major_fix_required
|
||||
```
|
||||
|
||||
### Example 4: Exploration
|
||||
**Review 完成度路由**:
|
||||
```
|
||||
User: 我想优化系统性能，但不知道从哪入手
|
||||
CCW: Intent=exploration, Workflow=brainstorm→plan→execute
|
||||
→ /workflow:brainstorm:auto-parallel "探索系统性能优化方向"
|
||||
critical == 0 AND high <= 3 → complete_or_optional_fix
|
||||
critical > 0 → mandatory_fix
|
||||
high > 3 → recommended_fix
|
||||
```
|
||||
|
||||
### Example 5: Multi-Model Collaboration
|
||||
### 流转决策模式
|
||||
|
||||
**plan_execute_test**:
|
||||
```
|
||||
User: 用 gemini 分析现有架构，然后让 codex 实现优化
|
||||
CCW: Detects tool preferences, executes in sequence
|
||||
→ Gemini CLI (analysis) → Codex CLI (implementation)
|
||||
plan → execute → test
|
||||
↓ (if test fail)
|
||||
extract_failures → fix → test (max 3 iterations)
|
||||
↓ (if still fail)
|
||||
manual_intervention
|
||||
```
|
||||
|
||||
**iterative_improvement**:
|
||||
```
|
||||
execute → test → fix → test → ...
|
||||
loop until: pass_rate >= 0.95 OR iterations >= 3
|
||||
```
|
||||
|
||||
### 使用示例
|
||||
|
||||
```javascript
|
||||
// 执行完成后，根据产出决定下一步
|
||||
const result = await execute(plan)
|
||||
|
||||
// 提取意图流转到测试
|
||||
const testContext = extractIntent('execute_to_test', result)
|
||||
// testContext = { modified_files, test_scope, pending_verification }
|
||||
|
||||
// 测试完成后，根据完成度决定路由
|
||||
const testResult = await test(testContext)
|
||||
const nextStep = evaluateCompletion('test', testResult)
|
||||
// nextStep = 'fix_failures_then_continue' if pass_rate = 0.85
|
||||
```
|
||||
|
||||
## Reference
|
||||
|
||||
- [command.json](command.json) - 命令元数据、Flow 定义、意图规则、Artifact Flow
|
||||
|
||||
Reference in New Issue
Block a user