apply

Tidy Apply

Apply the HIGH-confidence changes from a previous --dry-run invocation.

RUN_ID="$ARGUMENTS"

If $ARGUMENTS is empty, list recent runs with status inspection_complete and ask which to apply.

Steps

1. Verify .tidy/runs/$RUN_ID/state.json exists. If not, fail with helpful message.

2. Read state.json. Check status — must be inspection_complete or cancelled_at_review. If anything else, tell user the run is in an unexpected state (point them at the run dir).

3. Verify the working tree is still clean:

git status --porcelain

If dirty: refuse, instruct user to stash/commit/branch first.

1. Verify the pre_tidy_sha is still reachable (in case they merged or rebased):

git rev-parse "$PRE_TIDY_SHA" >/dev/null 2>&1

If not, the historical baseline is gone — tell the user the run cannot be applied because the baseline was rewritten. Suggest re-running /tidy:run --apply fresh.

1. Read each track's assessment.md. Present a brief summary to the user:

Tidy run <RUN_ID> inspection found:
- HIGH-confidence (will apply): <X> total across N tracks
- MEDIUM-confidence (will skip — needs review): <Y>
- LOW (will skip — flagged): <Z>

Last inspected: <created_at>
Tracks: <list>

Proceed with apply?

1. Call AskUserQuestion:

   • Apply — proceed
   • Review HIGH list first — render the HIGH-confidence items from each assessment, then re-ask
   • Cancel — stop, no changes made

2. On Apply: invoke the apply phase from commands/run.md PHASE 4 onward — same logic, same track-application order, same reviewer pass. The orchestrator code in run.md is the canonical implementation; this command is a re-entry point.

3. Update state.json: mode: "apply", applied_at: "<ISO>".

Rules

• Same refuse-on-dirty-tree rule as run.md PHASE 0.2.
• Same track-application order (deprecated-slop → dead-code → dedup → type-consolidate → type-strengthen → error-cleanup → circular-deps).
• Reviewer pass is mandatory.
• If the user runs apply on a completed run, refuse. They should run /tidy:run again for a fresh pass.