mimimaster/piwin/executing-plans
Claude Code and agent skill by mimimaster/piwin
Trust tier 3
- community
Source check pending
Checking that the SKILL.md file is still in the repository.
Execute an approved SessionPlan in the current session slice by slice, recording step evidence with piwin_plan_set_step. Use when the user wants an approved plan implemented inline (执行一下 / 当前会话执行 / 按计划做) or asks to continue the next plan step. If they chose 子代理执行, use subagent-driven-development instead.
Install
npx vskill@latest install mimimaster/piwin/executing-plansSKILL.md
Executing Plans
Goal
Implement only the current approved plan slice so its acceptance criteria pass, then stop at the next checkpoint.
Done means
- Current step(s) match plan intent; no drive-by refactors or scope expansion.
- Step progress is recorded with
piwin_plan_set_step(markdoneonly with a short evidence note). - Required verification for the slice ran in this environment (tests, typecheck, or the plan’s stated check).
- If a step has
profileId, the Host-resolved profile bounds model/capabilities/isolation — do not widen them.
Before you execute
- If the user already named a mode, or plan context already shows a chosen execution mode, use that path. Do not ask them to repeat keywords.
- If they only say they want it executed (for example 「执行一下」) and no mode is known, ask 当前会话执行,还是子代理执行? Then wait.
- If the plan is still draft, or
piwin_plan_set_stepsays 尚未选择执行方式 / 请先确认执行方式, ask that same question. Do not retry the tool and do not keep implementing as if the plan were approved. - Failed runs use the existing continue / retry controls. Do not invent a second recovery path.
Stop when
- Plan is wrong, blocked, or acceptance criteria cannot be met — fix the plan briefly or ask; do not silently diverge.
- Checkpoint / user gate requires review before the next slice.
Constraints
- Default profiles when unset: worktree + explicit apply. Built-ins:
explorer,reviewer(readonly);implementer,tester(worktree). - Host owns isolation and permission; skill text is guidance only.
- Worktree writes are single-threaded. Never assume two implementer children run at once.
Verify
- Re-run the failing or required path for the slice; summarize actual results before claiming the step done.
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/executing-plans/SKILL.md
Popularity
- 7-day trend score
- 5
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/mimimaster/piwin/executing-plans)