| name | execute |
| description | Enters execution mode to implement a plan step by step. Invoke with the plan file path or after /createplan is complete. Implements one step at a time, updates the plan tracker, and writes a status report after each step. |
| argument-hint | ["path/to/plan.md"] |
You are entering Execution Mode. Implement the plan precisely — nothing more, nothing less.
Before you write a single line
- Read the full plan. Not just the current step — all of it. Understand the shape of the whole before touching any part.
- Locate existing patterns. For every file you're about to touch or create, find the closest existing example in the codebase and follow it. Match the style exactly — naming conventions, file structure, export patterns, everything.
- Confirm you have what you need. If the plan references a file or function you can't find, stop and say so before writing anything.
How to implement
One step at a time. Implement step 1. Verify it works. Write the status report. Then — and only then — move to step 2.
For each step:
1. Implement
- Write the minimum code that fully delivers the step's goal. No extras.
- Follow existing patterns exactly — error handling, typing, naming, file organisation.
- If you notice something that could be improved but isn't in the plan: do not fix it. Add a
// NOTE: comment and include it in your status report.
2. Update the plan tracker
After implementing each step, update the plan markdown file:
- Change the step's emoji from 🟥 to 🟩
- Update subtask emojis as each subtask completes
- Recalculate and update the top-line
**Status:** percentage
3. Write a status report
End every step with a clearly formatted status report:
## Status Report — Step [N]: [Name]
**Result:** ✅ Complete / ⚠️ Complete with notes / ❌ Blocked
**Files changed:**
- `src/path/to/file.ts` — [what changed]
**Functions added / modified:**
- `functionName()` — [one-line description]
**Decisions made during implementation:**
- [Any choice not explicitly specified in the plan, with reasoning]
**Notes for review:**
- [Anything that looked off, fragile, or worth a second pair of eyes]
- [Out-of-scope improvements noticed but not touched]
**Ready for next step:** Yes / No — [if No, explain what's blocking]
Rules
- Never go off-plan. If something unexpected requires a decision not covered in the plan, stop and ask rather than guessing.
- No scope creep. If it's not in the plan, it doesn't get built. No "while I'm here" changes.
- Match existing error handling. Find how errors are handled elsewhere and do the same.
- Don't break what works. Before moving to the next step, verify what you just implemented doesn't break adjacent functionality. If tests exist, run them.
- Be explicit about uncertainty. If the plan says "update X" but you find two candidates for X, don't pick one silently. Flag it.
If you get stuck
If you hit something unexpected — stop immediately. Do not improvise a workaround. Write a status report with ❌ Blocked and explain exactly what you found and what you need to proceed.
