Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build or refactor React and TypeScript user interfaces in the front or mobile packages. Use for components, pages, interactions, state, accessibility, responsive layout, API integration, and frontend architecture. Do not use for server-only changes.
.claude/skills/qierkang-frontend-development/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 47% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -70% | 0% |
| case-02 | ✓→✗ | ▼ Worse | -91% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -77% | 0% |
Use this skill for work under omni-platform-front/ or omni-platform-mobile/.
package.json.references/component-checklist.md when component ownership, interaction, accessibility, or responsive behavior matters.karpathy-guidelines for scope control and verification discipline.Run the affected package build from the repository root. Use browser or screenshot verification for visual or interaction changes when available.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | pass→pass | 9,945 | 13,604 | +37% | 1 | 1 | 0% | 1,901 | 3,091 | +63% | 0 | 0 | — |
case-01 | fail→fail | 3,951 | 6,754 | +71% | 1 | 1 | 0% | 184 | 563 | +206% | 0 | 0 | — |
case-02 | pass→fail | 26,313 | 4,731 | -82% | 1 | 1 | 0% | 5,784 | 501 | -91% | 0 | 0 | — |
case-03 | fail→fail | 4,204 | 4,140 | -2% | 1 | 1 | 0% | 267 | 528 | +98% | 0 | 0 | — |
case-04 | pass→fail | 10,650 | 3,512 | -67% | 1 | 1 | 0% | 2,175 | 510 | -77% | 0 | 0 | — |
case-05 | pass→fail | 15,662 | 4,997 | -68% | 1 | 1 | 0% | 3,002 | 494 | -84% | 0 | 0 | — |
case-06 | pass→fail | 13,077 | 3,608 | -72% | 1 | 1 | 0% | 2,735 | 411 | -85% | 0 | 0 | — |
case-07 | pass→fail | 13,621 | 2,825 | -79% | 1 | 1 | 0% | 2,235 | 655 | -71% | 0 | 0 | — |
case-08 | fail→pass | 11,125 | 14,064 | +26% | 1 | 1 | 0% | 2,127 | 3,133 | +47% | 0 | 0 | — |
case-09 | fail→fail | 7,906 | 7,418 | -6% | 1 | 1 | 0% | 1,781 | 861 | -52% | 0 | 0 | — |
case-10 | fail→fail | 13,209 | 5,820 | -56% | 1 | 1 | 0% | 2,301 | 715 | -69% | 0 | 0 | — |
case-12 | fail→fail | 9,841 | 39,136 | +298% | 1 | 1 | 0% | 2,042 | 673 | -67% | 0 | 0 | — |
case-13 | pass→pass | 7,030 | 5,277 | -25% | 1 | 1 | 0% | 1,205 | 1,253 | +4% | 0 | 0 | — |
case-14 | pass→pass | 11,133 | 5,059 | -55% | 1 | 1 | 0% | 1,655 | 1,060 | -36% | 0 | 0 | — |
case-15 | fail→pass | 12,102 | 5,492 | -55% | 1 | 1 | 0% | 2,198 | 1,247 | -43% | 0 | 0 | — |
case-16 | fail→pass | 11,921 | 1,824 | -85% | 1 | 1 | 0% | 1,893 | 560 | -70% | 0 | 0 | — |
case-17 | fail→fail | 10,023 | 3,502 | -65% | 1 | 1 | 0% | 1,877 | 844 | -55% | 0 | 0 | — |
case-18 | pass→fail | 20,569 | 6,119 | -70% | 1 | 1 | 0% | 3,921 | 657 | -83% | 0 | 0 | — |
case-19 | pass→fail | 14,861 | 4,906 | -67% | 1 | 1 | 0% | 3,288 | 479 | -85% | 0 | 0 | — |
case-20 | fail→fail | 12,258 | 4,894 | -60% | 1 | 1 | 0% | 2,553 | 514 | -80% | 0 | 0 | — |
case-21 | pass→pass | 11,959 | 5,436 | -55% | 1 | 1 | 0% | 2,031 | 1,166 | -43% | 0 | 0 | — |
case-22 | fail→fail | 13,971 | 6,409 | -54% | 1 | 1 | 0% | 2,796 | 619 | -78% | 0 | 0 | — |
case-23 | pass→fail | 11,124 | 4,762 | -57% | 1 | 1 | 0% | 2,299 | 555 | -76% | 0 | 0 | — |
case-24 | fail→fail | 8,231 | 5,996 | -27% | 1 | 1 | 0% | 1,782 | 711 | -60% | 0 | 0 | — |
case-25 | pass→pass | 4,916 | 3,757 | -24% | 1 | 1 | 0% | 953 | 923 | -3% | 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. 25 cases were attempted, and 10 counted toward the lift figure. The other 15 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -20 percentage points is the difference between those two pass rates over the 10 comparable cases. 11 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.
Other measured skills in the registry, with their headline benchmark lift.