Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →기존 references/<id>의 §8 Component Patterns + frontmatter `tokens.components`를 멀티서피스(여러 라우트 + 메뉴/모달 인터랙션) + 공개 디자인시스템(Storybook/Primer/Polaris/Cedar/Geist/*.design) 크롤로 풍부화. 단일 랜딩 스냅샷의 '버튼 수준'을 모달·탭·테이블·토스트·폼상태까지 확장하되, 소스가 빈약하면(랜딩만 있는 앱 중심 기업) 억지로 만들지 않고 정직하게 cap. 완료 시 `tokens.components_harvested: true` 마커. '컴포넌트 풍부화', '컴포넌트 하베스트', 'X 컴포넌트 보강', 'component harvest' 류에 트리거. 토큰 자체가 없으면 먼저 omd-token-backfill/omd-add-reference.
.claude/skills/kwakseongjae-omd-component-harvest/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 14% | 0% |
CORE_V2_CATALOG_WRITE_BLOCKED
Do not rewrite a reference section, legacy frontmatter token block, mirror, or quality marker. The current public catalog is dual-read source material, while all future writers must be graph-backed Core v2. Reusing the retired section and frontmatter mutation path would create a new legacy authority and can silently lose provenance.
Report that component harvesting is temporarily read-only. Evidence collection may be proposed separately, but no catalog bytes may change until a canonical catalog graph/evidence writer, migrated readers, and a provider-free dropped_segments=0 migration gate ship together.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 4,021 | 28,045 | +597% | 1 | 1 | 0% | 766 | 678 | -11% | 0 | 0 | — |
case-02 | fail→pass | 3,597 | 2,542 | -29% | 1 | 1 | 0% | 542 | 673 | +24% | 0 | 0 | — |
case-03 | fail→pass | 28,742 | 3,281 | -89% | 1 | 1 | 0% | 1,776 | 766 | -57% | 0 | 0 | — |
case-04 | fail→pass | 8,555 | 3,417 | -60% | 1 | 1 | 0% | 1,526 | 827 | -46% | 0 | 0 | — |
case-05 | fail→pass | 3,692 | 2,555 | -31% | 1 | 1 | 0% | 572 | 651 | +14% | 0 | 0 | — |
case-06 | fail→pass | 15,431 | 3,805 | -75% | 1 | 1 | 0% | 2,281 | 872 | -62% | 0 | 0 | — |
case-07 | fail→pass | 2,360 | 3,447 | +46% | 1 | 1 | 0% | 336 | 634 | +89% | 0 | 0 | — |
case-08 | fail→pass | 19,579 | 3,190 | -84% | 1 | 1 | 0% | 3,887 | 702 | -82% | 0 | 0 | — |
case-09 | fail→pass | 12,350 | 3,781 | -69% | 1 | 1 | 0% | 2,405 | 851 | -65% | 0 | 0 | — |
case-10 | fail→pass | 2,322 | 3,312 | +43% | 1 | 1 | 0% | 365 | 705 | +93% | 0 | 0 | — |
case-11 | fail→pass | 3,070 | 2,750 | -10% | 1 | 1 | 0% | 510 | 700 | +37% | 0 | 0 | — |
case-12 | fail→pass | 13,983 | 33,181 | +137% | 1 | 1 | 0% | 2,102 | 623 | -70% | 0 | 0 | — |
case-13 | fail→pass | 12,378 | 3,638 | -71% | 1 | 1 | 0% | 2,646 | 812 | -69% | 0 | 0 | — |
case-14 | fail→pass | 13,908 | 2,926 | -79% | 1 | 1 | 0% | 2,906 | 695 | -76% | 0 | 0 | — |
case-15 | fail→pass | 12,283 | 3,852 | -69% | 1 | 1 | 0% | 2,190 | 865 | -61% | 0 | 0 | — |
case-16 | fail→pass | 3,899 | 3,523 | -10% | 1 | 1 | 0% | 624 | 712 | +14% | 0 | 0 | — |
case-17 | fail→pass | 21,290 | 3,959 | -81% | 1 | 1 | 0% | 4,735 | 792 | -83% | 0 | 0 | — |
case-18 | fail→pass | 11,785 | 4,134 | -65% | 1 | 1 | 0% | 2,440 | 772 | -68% | 0 | 0 | — |
case-19 | pass→pass | 18,546 | 11,499 | -38% | 1 | 1 | 0% | 3,343 | 2,066 | -38% | 0 | 0 | — |
case-20 | pass→pass | 10,677 | 8,059 | -25% | 1 | 1 | 0% | 1,822 | 1,661 | -9% | 0 | 0 | — |
case-21 | fail→fail | 32,103 | 4,167 | -87% | 1 | 1 | 0% | 4,733 | 981 | -79% | 0 | 0 | — |
case-22 | pass→pass | 20,382 | 18,839 | -8% | 1 | 1 | 0% | 4,033 | 3,297 | -18% | 0 | 0 | — |
case-23 | pass→fail | 15,108 | 7,358 | -51% | 1 | 1 | 0% | 3,331 | 1,593 | -52% | 0 | 0 | — |
case-24 | pass→fail | 18,499 | 9,025 | -51% | 1 | 1 | 0% | 3,404 | 1,798 | -47% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted. The headline lift of +67 percentage points is the difference between those two pass rates over the 24 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/20/2026 | +45% |
Other measured skills in the registry, with their headline benchmark lift.