do
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
Chinesedo - Feature Development Orchestrator
do - 功能开发编排器
An orchestrator for systematic feature development. Invoke agents via , never write code directly.
codeagent-wrapper一个用于系统化功能开发的编排器。通过调用agents,切勿直接编写代码。
codeagent-wrapperLoop Initialization (REQUIRED)
循环初始化(必填)
When triggered via , follow these steps:
/do <task>当通过触发时,请遵循以下步骤:
/do <task>Step 1: Ask about worktree mode
步骤1:询问工作树模式
Use AskUserQuestion to ask:
Develop in a separate worktree? (Isolates changes from main branch)
- Yes (Recommended for larger changes)
- No (Work directly in current directory)使用AskUserQuestion询问:
是否在独立工作树中开发?(将变更与主分支隔离)
- 是(推荐用于较大变更)
- 否(直接在当前目录工作)Step 2: Initialize state
步骤2:初始化状态
bash
undefinedbash
undefinedIf worktree mode selected:
若选择工作树模式:
python3 "${SKILL_DIR}/scripts/setup-do.py" --worktree "<task description>"
python3 "${SKILL_DIR}/scripts/setup-do.py" --worktree "<task description>"
If no worktree:
若不使用工作树:
python3 "${SKILL_DIR}/scripts/setup-do.py" "<task description>"
This creates `.claude/do.{task_id}.local.md` with:
- `active: true`
- `current_phase: 1`
- `max_phases: 5`
- `completion_promise: "<promise>DO_COMPLETE</promise>"`
- `use_worktree: true/false`python3 "${SKILL_DIR}/scripts/setup-do.py" "<task description>"
此命令会创建`.claude/do.{task_id}.local.md`文件,包含以下内容:
- `active: true`
- `current_phase: 1`
- `max_phases: 5`
- `completion_promise: "<promise>DO_COMPLETE</promise>"`
- `use_worktree: true/false`Worktree Mode
工作树模式
When in state file, ALL calls that modify code MUST include :
use_worktree: truecodeagent-wrapper--worktreebash
undefined当状态文件中时,所有会修改代码的调用必须包含参数:
use_worktree: truecodeagent-wrapper--worktreebash
undefinedWith worktree mode enabled
启用工作树模式时
codeagent-wrapper --worktree --agent develop - . <<'EOF'
...
EOF
codeagent-wrapper --worktree --agent develop - . <<'EOF'
...
EOF
Parallel tasks with worktree
工作树模式下的并行任务
codeagent-wrapper --worktree --parallel <<'EOF'
---TASK---
id: task1
agent: develop
workdir: .
---CONTENT---
...
EOF
The `--worktree` flag tells codeagent-wrapper to create/use a worktree internally. Read-only agents (code-explorer, code-architect, code-reviewer) do NOT need `--worktree`.codeagent-wrapper --worktree --parallel <<'EOF'
---TASK---
id: task1
agent: develop
workdir: .
---CONTENT---
...
EOF
`--worktree`参数用于告知codeagent-wrapper在内部创建/使用工作树。只读agents(code-explorer、code-architect、code-reviewer)无需使用`--worktree`。Loop State Management
循环状态管理
After each phase, update frontmatter:
.claude/do.{task_id}.local.mdyaml
current_phase: <next phase number>
phase_name: "<next phase name>"When all 5 phases complete, output the completion signal:
<promise>DO_COMPLETE</promise>To abort early, set in the state file.
active: false每个阶段完成后,更新的前置内容:
.claude/do.{task_id}.local.mdyaml
current_phase: <下一阶段编号>
phase_name: "<下一阶段名称>"当所有5个阶段完成后,输出完成信号:
<promise>DO_COMPLETE</promise>若要提前终止,将状态文件中的。
active: falseHard Constraints
硬性约束
- Never write code directly. Delegate all code changes to agents.
codeagent-wrapper - Pass complete context forward. Every agent invocation includes the Context Pack.
- Parallel-first. Run independent tasks via .
codeagent-wrapper --parallel - Update state after each phase. Keep current.
.claude/do.{task_id}.local.md - Expect long-running calls. High-reasoning modes can take a long time; stay in the orchestrator role and wait for agents to complete.
codeagent-wrapper - Timeouts are not an escape hatch. If a invocation times out/errors, retry (split/narrow the task if needed); never switch to direct implementation.
codeagent-wrapper - Respect worktree setting. If , always pass
use_worktree: trueto develop agent calls.--worktree
- 切勿直接编写代码。 所有代码变更都需委托给agents处理。
codeagent-wrapper - 完整传递上下文。 每次调用agent时都需包含上下文包。
- 优先并行处理。 通过运行独立任务。
codeagent-wrapper --parallel - 每个阶段后更新状态。 保持为最新状态。
.claude/do.{task_id}.local.md - codeagent-wrapper调用可能耗时较长。 高推理模式可能需要较长时间;请保持编排器角色,等待agents完成任务。
- 超时不是逃避手段。 若调用超时/出错,请重试(必要时拆分/缩小任务范围);切勿切换为直接实现。
codeagent-wrapper - 遵守工作树设置。 若,调用develop agent时务必传递
use_worktree: true参数。--worktree
Agents
Agents
| Agent | Purpose | Needs --worktree |
|---|---|---|
| Trace code, map architecture, find patterns | No (read-only) |
| Design approaches, file plans, build sequences | No (read-only) |
| Review for bugs, simplicity, conventions | No (read-only) |
| Implement code, run tests | Yes (if worktree enabled) |
| Agent | 用途 | 是否需要--worktree |
|---|---|---|
| 追踪代码、绘制架构图、发现模式 | 否(只读) |
| 设计实现方案、文件规划、构建序列 | 否(只读) |
| 检查Bug、简洁性、编码规范 | 否(只读) |
| 实现代码、运行测试 | 是(若启用工作树) |
Issue Severity Definitions
问题严重程度定义
Blocking issues (require user input):
- Impacts core functionality or correctness
- Security vulnerabilities
- Architectural conflicts with existing patterns
- Ambiguous requirements with multiple valid interpretations
Minor issues (auto-fix without asking):
- Code style inconsistencies
- Naming improvements
- Missing documentation
- Non-critical test coverage gaps
阻塞性问题(需要用户输入):
- 影响核心功能或正确性
- 安全漏洞
- 与现有模式存在架构冲突
- 需求模糊,存在多种有效解读
次要问题(无需询问,自动修复):
- 代码风格不一致
- 命名优化
- 缺少文档
- 非关键测试覆盖缺口
Context Pack Template
上下文包模板
text
undefinedtext
undefinedOriginal User Request
原始用户请求
<verbatim request>
<原始请求内容>
Context Pack
上下文包
- Phase: <1-5 name>
- Decisions: <requirements/constraints/choices>
- Code-explorer output: <paste or "None">
- Code-architect output: <paste or "None">
- Code-reviewer output: <paste or "None">
- Develop output: <paste or "None">
- Open questions: <list or "None">
- 阶段:<1-5 阶段名称>
- 决策:<需求/约束/选择>
- Code-explorer输出:<粘贴内容或"无">
- Code-architect输出:<粘贴内容或"无">
- Code-reviewer输出:<粘贴内容或"无">
- Develop输出:<粘贴内容或"无">
- 待解决问题:<列表或"无">
Current Task
当前任务
<specific task>
<具体任务>
Acceptance Criteria
验收标准
<checkable outputs>
```
<可检查的输出项>
undefined5-Phase Workflow
5阶段工作流
Phase 1: Understand (Parallel, No Interaction)
阶段1:理解(并行,无需交互)
Goal: Understand requirements and map codebase simultaneously.
Actions: Run and 2-3 tasks in parallel.
code-architectcode-explorerbash
codeagent-wrapper --parallel <<'EOF'
---TASK---
id: p1_requirements
agent: code-architect
workdir: .
---CONTENT---目标: 同时理解需求并绘制代码库映射。
操作: 并行运行和2-3个任务。
code-architectcode-explorerbash
codeagent-wrapper --parallel <<'EOF'
---TASK---
id: p1_requirements
agent: code-architect
workdir: .
---CONTENT---Original User Request
原始用户请求
/do <request>
/do <request>
Context Pack
上下文包
- Code-explorer output: None
- Code-architect output: None
- Code-explorer输出:无
- Code-architect输出:无
Current Task
当前任务
- Analyze requirements completeness (score 1-10)
- Extract explicit requirements, constraints, acceptance criteria
- Identify blocking questions (issues that prevent implementation)
- Identify minor clarifications (nice-to-have but can proceed without)
Output format:
- Completeness score: X/10
- Requirements: [list]
- Non-goals: [list]
- Blocking questions: [list, if any]
- Minor clarifications: [list, if any]
- 分析需求完整性(评分1-10)
- 提取明确需求、约束条件、验收标准
- 识别阻塞性问题(阻碍实现的问题)
- 识别次要澄清点(有则更好,但不影响推进)
输出格式:
- 完整性评分:X/10
- 需求:[列表]
- 非目标:[列表]
- 阻塞性问题:[列表,如有]
- 次要澄清点:[列表,如有]
Acceptance Criteria
验收标准
Concrete checklist; blocking vs minor questions clearly separated.
---TASK---
id: p1_similar_features
agent: code-explorer
workdir: .
---CONTENT---
具体检查清单;明确区分阻塞性与次要问题。
---TASK---
id: p1_similar_features
agent: code-explorer
workdir: .
---CONTENT---
Original User Request
原始用户请求
/do <request>
/do <request>
Current Task
当前任务
Find 1-3 similar features, trace end-to-end. Return: key files with line numbers, call flow, extension points.
查找1-3个类似功能,进行端到端追踪。返回:带行号的关键文件、调用流程、扩展点。
Acceptance Criteria
验收标准
Concrete file:line map + reuse points.
---TASK---
id: p1_architecture
agent: code-explorer
workdir: .
---CONTENT---
具体的文件:行号映射 + 可复用点。
---TASK---
id: p1_architecture
agent: code-explorer
workdir: .
---CONTENT---
Original User Request
原始用户请求
/do <request>
/do <request>
Current Task
当前任务
Map architecture for relevant subsystem. Return: module map + 5-10 key files.
绘制相关子系统的架构图。返回:模块映射 + 5-10个关键文件。
Acceptance Criteria
验收标准
Clear boundaries; file:line references.
---TASK---
id: p1_conventions
agent: code-explorer
workdir: .
---CONTENT---
清晰的边界;包含文件:行号引用。
---TASK---
id: p1_conventions
agent: code-explorer
workdir: .
---CONTENT---
Original User Request
原始用户请求
/do <request>
/do <request>
Current Task
当前任务
Identify testing patterns, conventions, config. Return: test commands + file locations.
识别测试模式、编码规范、配置。返回:测试命令 + 文件位置。
Acceptance Criteria
验收标准
Test commands + relevant test file paths.
EOF
undefined测试命令 + 相关测试文件路径。
EOF
undefinedPhase 2: Clarify (Conditional)
阶段2:澄清(条件性步骤)
Goal: Resolve blocking ambiguities only.
Actions:
- Review output for blocking questions
p1_requirements - IF blocking questions exist → Use AskUserQuestion
- IF no blocking questions (completeness >= 8) → Skip to Phase 3, log "Requirements clear, proceeding"
bash
undefined目标: 仅解决阻塞性歧义。
操作:
- 查看输出中的阻塞性问题
p1_requirements - 若存在阻塞性问题 → 使用AskUserQuestion询问用户
- 若没有阻塞性问题(完整性评分≥8) → 跳过至阶段3,记录“需求明确,继续推进”
bash
undefinedOnly if blocking questions exist:
仅当存在阻塞性问题时执行:
Use AskUserQuestion with the blocking questions from Phase 1
使用AskUserQuestion询问阶段1中识别的阻塞性问题
undefinedundefinedPhase 3: Design (No Interaction)
阶段3:设计(无需交互)
Goal: Produce minimal-change implementation plan.
Actions: Invoke with all Phase 1 context to generate a single implementation plan.
code-architectbash
codeagent-wrapper --agent code-architect - . <<'EOF'目标: 生成最小变更的实现方案。
操作: 调用并传入阶段1的所有上下文,生成单一实现方案。
code-architectbash
codeagent-wrapper --agent code-architect - . <<'EOF'Original User Request
原始用户请求
/do <request>
/do <request>
Context Pack
上下文包
- Code-explorer output: <ALL Phase 1 explorer outputs>
- Code-architect output: <Phase 1 requirements + Phase 2 answers if any>
- Code-explorer输出:<阶段1所有explorer输出>
- Code-architect输出:<阶段1需求 + 阶段2答案(如有)>
Current Task
当前任务
Design minimal-change implementation:
- Reuse existing abstractions
- Minimize new files
- Follow established patterns from code-explorer output
Output:
- File touch list with specific changes
- Build sequence
- Test plan
- Risks and mitigations
设计最小变更的实现方案:
- 复用现有抽象
- 最小化新增文件
- 遵循code-explorer输出中的既定模式
输出:
- 需修改的文件列表及具体变更内容
- 构建序列
- 测试方案
- 风险及缓解措施
Acceptance Criteria
验收标准
Concrete, implementable blueprint with minimal moving parts.
EOF
undefined具体、可实现的蓝图,尽可能减少变动点。
EOF
undefinedPhase 4: Implement + Review (Single Interaction Point)
阶段4:实现 + 评审(单一交互点)
Goal: Build feature and review in one phase.
Actions:
- Invoke to implement (add
developif--worktree):use_worktree: true
bash
undefined目标: 完成功能构建并在本阶段内完成评审。
操作:
- 调用进行实现(若
develop则添加use_worktree: true参数):--worktree
bash
undefinedCheck use_worktree from state file, add --worktree if true
检查状态文件中的use_worktree设置,若为true则添加--worktree参数
codeagent-wrapper --worktree --agent develop - . <<'EOF'
codeagent-wrapper --worktree --agent develop - . <<'EOF'
Original User Request
原始用户请求
/do <request>
/do <request>
Context Pack
上下文包
- Code-explorer output: <ALL Phase 1 outputs>
- Code-architect output: <Phase 3 blueprint>
- Code-explorer输出:<阶段1所有输出>
- Code-architect输出:<阶段3蓝图>
Current Task
当前任务
Implement with minimal change set following the blueprint.
- Follow Phase 1 patterns
- Add/adjust tests per Phase 3 plan
- Run narrowest relevant tests
遵循蓝图,以最小变更集实现功能。
- 遵循阶段1识别的模式
- 根据阶段3的测试方案添加/调整测试
- 运行最相关的窄范围测试
Acceptance Criteria
验收标准
Feature works end-to-end; tests pass; diff is minimal.
EOF
2. Run parallel reviews (no --worktree needed, read-only):
```bash
codeagent-wrapper --parallel <<'EOF'
---TASK---
id: p4_correctness
agent: code-reviewer
workdir: .
---CONTENT---功能端到端可用;测试通过;代码差异最小。
EOF
2. 并行运行评审任务(无需--worktree,只读):
```bash
codeagent-wrapper --parallel <<'EOF'
---TASK---
id: p4_correctness
agent: code-reviewer
workdir: .
---CONTENT---Original User Request
原始用户请求
/do <request>
/do <request>
Context Pack
上下文包
- Code-architect output: <Phase 3 blueprint>
- Develop output: <implementation output>
- Code-architect输出:<阶段3蓝图>
- Develop输出:<实现输出>
Current Task
当前任务
Review for correctness, edge cases, failure modes.
Classify each issue as BLOCKING or MINOR.
评审正确性、边缘情况、故障模式。
将每个问题分类为阻塞性或次要问题。
Acceptance Criteria
验收标准
Issues with file:line references, severity, and concrete fixes.
---TASK---
id: p4_simplicity
agent: code-reviewer
workdir: .
---CONTENT---
问题需包含文件:行号引用、严重程度及具体修复方案。
---TASK---
id: p4_simplicity
agent: code-reviewer
workdir: .
---CONTENT---
Original User Request
原始用户请求
/do <request>
/do <request>
Context Pack
上下文包
- Code-architect output: <Phase 3 blueprint>
- Develop output: <implementation output>
- Code-architect输出:<阶段3蓝图>
- Develop输出:<实现输出>
Current Task
当前任务
Review for KISS: remove bloat, collapse needless abstractions.
Classify each issue as BLOCKING or MINOR.
基于KISS原则评审:移除冗余代码、合并不必要的抽象。
将每个问题分类为阻塞性或次要问题。
Acceptance Criteria
验收标准
Actionable simplifications with severity and justification.
EOF
3. Handle review results:
- **MINOR issues only** → Auto-fix via `develop` (with `--worktree` if enabled), no user interaction
- **BLOCKING issues** → Use AskUserQuestion: "Fix now / Proceed as-is"可操作的简化建议,包含严重程度及理由。
EOF
3. 处理评审结果:
- **仅存在次要问题** → 通过`develop`自动修复(若启用工作树则添加`--worktree`),无需用户交互
- **存在阻塞性问题** → 使用AskUserQuestion询问:“立即修复/保持现状继续推进”Phase 5: Complete (No Interaction)
阶段5:完成(无需交互)
Goal: Document what was built.
Actions: Invoke to produce summary:
code-reviewerbash
codeagent-wrapper --agent code-reviewer - . <<'EOF'目标: 记录已构建的内容。
操作: 调用生成总结:
code-reviewerbash
codeagent-wrapper --agent code-reviewer - . <<'EOF'Original User Request
原始用户请求
/do <request>
/do <request>
Context Pack
上下文包
- Code-architect output: <Phase 3 blueprint>
- Code-reviewer output: <Phase 4 review outcomes>
- Develop output: <Phase 4 implementation + fixes>
- Code-architect输出:<阶段3蓝图>
- Code-reviewer输出:<阶段4评审结果>
- Develop输出:<阶段4实现 + 修复内容>
Current Task
当前任务
Write completion summary:
- What was built
- Key decisions/tradeoffs
- Files modified (paths)
- How to verify (commands)
- Follow-ups (optional)
编写完成总结:
- 已构建的内容
- 关键决策/权衡
- 修改的文件(路径)
- 验证方法(命令)
- 后续工作(可选)
Acceptance Criteria
验收标准
Short, technical, actionable summary.
EOF
Output the completion signal:<promise>DO_COMPLETE</promise>
undefined简短、技术化、可执行的总结。
EOF
输出完成信号:<promise>DO_COMPLETE</promise>
undefined