con un clic
execute-plan
Use when you have a written implementation plan to execute
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Menú
Use when you have a written implementation plan to execute
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
Use when a plan's PR is ready to merge — marks the plan completed, checks PR status, merges to main, and syncs local main
Use when a feature or change needs an implementation plan with concrete tasks, file paths, and test steps before coding begins
Provides comprehensive code review guidance for React 19, Vue 3, Rust, TypeScript, Java, Python, C/C++, Go, and WASM. Helps catch bugs, improve code quality, and give constructive feedback through systematic analysis. Features detailed language-specific guides, severity classification, review techniques, and hard-stop protocol. Use when: reviewing pull requests, conducting PR reviews, code review, reviewing code changes, establishing review standards, mentoring developers, architecture reviews, security audits, checking code quality, finding bugs, giving feedback on code.
Use before any creative work - creating features, building components, or modifying behavior
Use when creating new skills, editing existing skills, or verifying skills work before deployment
| name | execute-plan |
| description | Use when you have a written implementation plan to execute |
Read the plan, create a feature branch, dispatch subagents per task, review the branch, open a PR.
gitflow-branching skill if needed)manage_todo_list with all tasks from the planFor each task, dispatch a general subagent (sequentially, not parallel):
Subagent prompt template:
description: "Implement Task N: [task name]"
prompt: |
You are BUILDING code in [project]. DO NOT PLAN — write actual file edits using the Edit and Write tools.
## CRITICAL INSTRUCTIONS
1. READ the files listed below first to understand the existing patterns
2. USE the Edit tool to modify existing files — do NOT rewrite entire files
3. USE the Write tool to create new files
4. RUN `cargo fmt --all`, then `cargo check --workspace`, then `cargo test --workspace` to validate
5. COMMIT your work with: `git commit -am "[commit message]"`
6. If cargo check fails, READ the error, READ the relevant source file, FIX the code with Edit tool, then re-run check
7. DO NOT output planning text — output actual file changes
## Task
[FULL TEXT of task from plan - paste it, don't make subagent read file]
## Context
[Where this fits, dependencies, what's already done]
## Instructions
- Implement exactly what the task specifies
- Write tests (TDD: failing test first, then implementation)
- Validate your work by running each step **independently and in order** — wait for each to finish before starting the next:
1. Formatting (e.g. `cargo fmt`, `prettier`, etc.)
2. Build / compile (e.g. `cargo build`, `npm run build`, etc.)
3. Tests (e.g. `cargo test`, `npm test`, etc.)
- If any step fails: **STOP. Do not re-run it yet.** Read the error output, read the relevant source files, then use Edit/Write tools to fix the root cause. Only re-run after you have made file changes.
- **Loop-break rule:** If you run a step and it fails, and you have made no file edits since the last time it failed, you are looping. Stop immediately and report back with status BLOCKED — describe the error and what you tried.
- Commit your work with a descriptive message
- Work from: [directory]
## Report back with:
- Status: DONE | BLOCKED | NEEDS_CONTEXT
- What you implemented
- Files changed
- Any concerns
Handle subagent responses:
Important: Dispatch tasks sequentially (not in parallel) to avoid file conflicts.
Once all tasks are done, ask the user what to do next:
ask({
questions: [{
id: "next-step",
question: "All tasks complete. What would you like to do next?",
options: [
{ label: "Code review then PR" },
{ label: "Open PR only" },
{ label: "Code review only" },
{ label: "Finish plan" }
]
}]
})
Then follow the user's choice immediately — do NOT ask for additional confirmation.
Dispatch the reviewer subagent with: "Review the implementation against the plan at docs/plans/YYYY-MM-DD-<feature>.md. Check that all acceptance criteria are met and no planned work was missed."
Fix any critical/major issues from reviewer verdict
Load the review skill to conduct a thorough code review
Clear the todo list — remove all old task entries
Create new todos for each finding from the review:
[severity] <brief description>Fix issues ONE AT A TIME using general subagents:
FOR EACH TODO ITEM, YOU MUST DISPATCH A SUBAGENT:
subagent({
agent: "general",
task: "You are BUILDING code in [project]. DO NOT PLAN — write actual file edits using the Edit and Write tools.\n\n## CRITICAL INSTRUCTIONS\n1. READ the files listed below first to understand the existing patterns\n2. USE the Edit tool to modify existing files — do NOT rewrite entire files\n3. RUN `cargo fmt --all`, then `cargo check --workspace`, then `cargo test --workspace` to validate\n4. COMMIT your work with: `git commit -am \"[commit message]\"`\n5. If cargo check fails, READ the error, READ the relevant source file, FIX the code with Edit tool, then re-run check\n6. DO NOT output planning text — output actual file changes\n\n## Issue to Fix\n[FULL TEXT from the todo item]\n\n## Instructions\n- Load the `review` skill for guidance on best practices\n- Fix the issue exactly as described\n- Validate your fix by running tests/linting/build\n- Update the corresponding todo in the todo list to \"completed\" using manage_todo_list\n- Commit your fix with a descriptive message\n- Work from: [directory]\n\n## Report back with:\n- Status: DONE | BLOCKED\n- What you fixed\n- Files changed",
description: "Fix: [severity] <brief description>"
})
CRITICAL: Do NOT fix the issue yourself. You MUST dispatch a general subagent for each todo item. Wait for the subagent to complete, then mark that todo as completed.
Re-run the review once after all fixes
If issues persist, escalate to user
Then proceed to Open PR below
Dispatch the reviewer subagent with: "Review the implementation against the plan at docs/plans/YYYY-MM-DD-<feature>.md. Check that all acceptance criteria are met and no planned work was missed."
Fix any critical/major issues from reviewer verdict
Load the review skill to conduct a thorough code review
Clear the todo list — remove all old task entries
Create new todos for each finding from the review:
[severity] <brief description>Fix issues ONE AT A TIME using general subagents:
FOR EACH TODO ITEM, YOU MUST DISPATCH A SUBAGENT:
subagent({
agent: "general",
task: "You are BUILDING code in [project]. DO NOT PLAN — write actual file edits using the Edit and Write tools.\n\n## CRITICAL INSTRUCTIONS\n1. READ the files listed below first to understand the existing patterns\n2. USE the Edit tool to modify existing files — do NOT rewrite entire files\n3. RUN `cargo fmt --all`, then `cargo check --workspace`, then `cargo test --workspace` to validate\n4. COMMIT your work with: `git commit -am \"[commit message]\"`\n5. If cargo check fails, READ the error, READ the relevant source file, FIX the code with Edit tool, then re-run check\n6. DO NOT output planning text — output actual file changes\n\n## Issue to Fix\n[FULL TEXT from the todo item]\n\n## Instructions\n- Load the `review` skill for guidance on best practices\n- Fix the issue exactly as described\n- Validate your fix by running tests/linting/build\n- Update the corresponding todo in the todo list to \"completed\" using manage_todo_list\n- Commit your fix with a descriptive message\n- Work from: [directory]\n\n## Report back with:\n- Status: DONE | BLOCKED\n- What you fixed\n- Files changed",
description: "Fix: [severity] <brief description>"
})
CRITICAL: Do NOT fix the issue yourself. You MUST dispatch a general subagent for each todo item.
Re-run the review once after all fixes
If issues persist, escalate to user
git push -u origin [branch-name]
gh pr create --title "[title]" --body "$(cat <<'EOF'
## Summary
- [bullets]
## Test plan
- [ ] [verification steps]
EOF
)"
Report the PR URL to the user.
Clear the todo list — remove all remaining entries now that execution is complete.
finish-plan skill to check PR status, merge to main, and update the plan indexAfter a PR is opened, update docs/plans/README.md:
docs: mark [plan-name] as completed