| name | work |
| description | Execute an approved plan step-by-step, with the repo's gate between steps and no scope creep. Use when implementing a planned change, working a checklist, or the user says "go" / "implement it" / "do the plan". Stops and reports on the first red gate. |
| argument-hint | [plan, or 'continue'] |
| allowed-tools | Read, Grep, Glob, Bash, Write, Edit |
Work the plan
Implement one step at a time. The point is a green, reviewable change — not speed.
If there's no plan yet and the change is non-trivial, run /spec first.
The loop (per step)
- State the step in one line so progress is visible.
- Build it in the surrounding style. If you're unsure a function, option, or
flag exists — yours or a dependency's —
/verify-api before you write it; don't
guess. Obey AGENTS.md and match .agent/rules/.
- Gate before moving on — run the repo's exact gate (
AGENTS.md → "The gate").
- Red gate → stop. Don't pile the next step on a broken one. Fix it, or report
the blocker with the error and your read on it. Never edit a test to make a real
failure pass.
Rules while working
- No scope creep. Build the approved slice. A good idea that wasn't in the plan
→
.agent/BACKLOG.md as "later", don't build it now. If the plan turns out wrong,
stop and re-plan with the user — don't silently redesign.
- Readable, no bloat. Match the surrounding style. Delete dead code you pass.
No speculative options or abstractions. Comments say why, not what.
- Tests are part of the step, not a follow-up — including the failure/denial path.
- Greenfield. Replacing code → delete the old and update every caller in the
same change; no shims for behavior nothing depends on yet.
When the change is green and reviewable, hand back for review (or, in a /sweep
run, self-review the diff, commit, and tick the task).
