zxdxjtu/harness/decompose
Claude Code and agent skill by zxdxjtu/harness
- community
Checking that the SKILL.md file is still in the repository.
Decompose tests into atomic task DAG with wave-based parallel execution and module stewardship
Install
npx vskill@latest install zxdxjtu/harness/decomposeSKILL.md
Decompose — Atomic Task DAG + Module Stewardship
Parameter: $ARGUMENTS — feature ID (e.g. F001)
Precondition
- Spec must be
approved(orreviewedif spec-review was in the flow) - Test files must exist and ALL be RED
Step 1: Analyze Test Cases
Read all test files for this feature. List every test ID.
Step 2: Build Atomic Tasks
Map each test ID (or group of closely related tests) to one atomic task.
Each task must satisfy:
- ≤ 2 hours of agent work
- One clear completion criterion: corresponding tests go from RED → GREEN
- Explicit inputs: which task outputs (files, interfaces) are required
- Explicit outputs: which files are produced + which tests turn green
- Independently verifiable: can run the test in isolation
Step 3: Analyze Dependencies
Determine task ordering:
- Type/interface definitions → implementations using those types
- Data layer → business logic → UI layer
- Unit test tasks → integration test tasks → E2E test tasks
Step 4: Assign Waves
- No dependencies → Wave 1 (parallel)
- Depends on Wave 1 → Wave 2 (parallel)
- ... and so on
Step 5: Module Stewardship Analysis
Analyze which code modules each task will touch:
- Identify module boundaries: Based on directory structure, package boundaries, or
.harness/full-design.md - Map tasks to modules: For each task, determine which module(s) it affects
- Generate AGENT.md files: For each affected module that doesn't already have one:
Read the template from the harness plugin's templates/agent-md-template.md and generate a module-specific AGENT.md:
Place AGENT.md in the actual module directory (e.g. src/auth/AGENT.md).
- Generate module dependency graph: Write
.harness/module-graph.json:
- Add stewardship metadata to tasks: Each task entry includes which module(s) it touches.
Step 6: Write tasks.md
Write to .harness/tasks.md:
Step 7: Update Progress
Append to .harness/progress.md:
Step 8: Request Approval
Present the DAG and module stewardship plan to the user. Ask:
"任务拆解完成:{N} 个原子任务,分 {M} 个 Wave,涉及 {K} 个模块(已生成 AGENT.md)。任务依赖关系是否正确?有需要拆分或合并的任务吗?"
Do NOT proceed until approved.
Next Step — Auto-Navigate
- Read the
flowfield from the spec - Display progress visualization:
- Ask: "任务拆解完成,共 {N} 个原子任务分 {M} 个 Wave。下一步是 Sprint 自动执行(按 Wave 并行开发,每 Wave 后触发模块守护和评估检查)。是否继续?"
- If user confirms → execute sprint logic directly
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/decompose/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/zxdxjtu/harness/decompose)