Polish

Skill jazz1x/galmuri/skills/polish

갈무리 — the Korean art of putting things away properly. A Claude Code plugin that gathers scattered thoughts into something you'd want to keep.

Install

npx -y skills add jazz1x/galmuri --skill polish

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Completion-claim gate. Before any "completed" / "done" / "passed" claim, output the self-check table. If ANY row is ✗, fix it first — do NOT report completion. Lean discipline-gate version (not the full quality-loop polish). Triggers: "polish", "완료 점검", "done check", "ready to ship", "PR ready", "완료 선언 전"

SKILL.md

4.8 KB, as published. Nobody here has run it

galmuri:polish — completion-claim gate

"검증 없이 완료 선언하지 않는다." A short, hard gate to run before saying done / 완료 / passed. If any row is ✗, fix it; do not declare completion.

If invoked bare (no scope hint), ask: "What are you about to declare complete? e.g. feat/foo branch, UI refactor, migration 0042 — name the scope so the checklist evidence makes sense."

This is the discipline-gate version. For the full quality-improvement loop (6-axis inspection × N + e2e + PR doc), use oh-my-borory:polish.

Step 1: Gather context

Cheap reads only — no full-file scans.

git status --short
git diff --stat
git branch --show-current
git log -1 --pretty='%h %s'

Note the baseline branch + SHA explicitly — the checklist below references it.

Step 2: Run the Self-Check Gate

Output exactly the following table. Every row must resolve to ✓ or ✗ — no n/a, no pending. If a row genuinely does not apply (e.g. no UI changes → testid row), mark it ✓ with evidence: scope-not-applicable.

| Check                                                          | Status | Evidence                                  |
|----------------------------------------------------------------|--------|-------------------------------------------|
| Branch is feature/* (not main/master/release/*)                | ✓/✗    | `git branch --show-current`               |
| Pre-commit hooks ran successfully                              | ✓/✗    | last hook exit code (or `.husky/pre-commit` ran) |
| Baseline branch + SHA explicitly stated for comparisons        | ✓/✗    | quote the SHA                             |
| No `withTimeout`, defensive `{ timeout }`, or silent try/catch | ✓/✗    | `rg 'withTimeout\|timeout:\s*\d+' diff`   |
| No hardcoded version field names (`v1.*`, `schema_v2`)         | ✓/✗    | `rg 'v\d+\.\w+\|schema_v\d' diff`         |
| testid constants are UPPER_SNAKE_CASE                          | ✓/✗    | `rg 'data-testid="[a-z]' diff`            |
| No raw px values outside design tokens                         | ✓/✗    | manual diff inspection / tokens file      |
| Long reports (>30 lines) written to docs/reports/ not chat     | ✓/✗    | inspect last chat output                  |
| Pixel-delta table present when claim is UI work                | ✓/✗    | path to pixel-delta artifact (or scope-not-applicable) |
| Locale separation respected (en/ko/jp files do not mix)        | ✓/✗    | `rg -l '[가-힣]\|[ぁ-ヿ]' *.md` filtered  |

Token Efficiency

After the table:

If the cycle exceeded 80% of context budget, write a handoff note to tmp/polish-handoff-<ts>.md and STOP. Wait for an explicit "continue".
Never restate the entire diff in chat — summarize and link to the commit hash.

Step 3: Halt-or-Pass

Any ✗ → halt. List each ✗ row with the fix command, apply fixes, re-run Step 2. Do not emit a completion claim until all rows are ✓.
All ✓ → emit one line: polish gate clean: {commit-hash} ready. That is the only form of completion claim allowed.

What this skill does NOT do

✗ Run inspections / write tests / generate PR docs (that's oh-my-borory:polish or harnish:ralphi)
✗ Modify any code by itself — it only verifies what you already did
✗ Replace galmuri:self-audit (which audits Claude's own session actions across all axes, not just completion claims)