Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Optimize UI performance against Core Web Vitals — LCP, INP, CLS — with loading/code-split strategy, layout-shift prevention, and animation performance rules. Use when the user wants to improve speed, fix jank or layout shift, hit Web Vitals budgets, or make a UI feel fast on low-end devices.
.claude/skills/plugin87-performance/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 73% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -2% | 0% |
Make the UI fast and stable. Treat performance as an accessibility concern — slow/janky UIs fail low-end devices first.
workflows/performance.md (Core Web Vitals targets, loading strategy, layout-shift, animation perf, design-system runtime cost).aspect-ratio (tokens/sizing.json), never inject content above existing.transform/opacity, will-change sparingly, 100–300ms (tokens/motion.json), honor prefers-reduced-motion. Prefer CSS state styling over JS for low INP; tree-shake/per-component imports.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 19,323 | 26,547 | +37% | 1 | 1 | 0% | 3,223 | 5,586 | +73% | 0 | 0 | — |
case-02 | fail→pass | 18,635 | 20,948 | +12% | 1 | 1 | 0% | 3,256 | 4,148 | +27% | 0 | 0 | — |
case-03 | fail→pass | 16,069 | 16,219 | +1% | 1 | 1 | 0% | 2,739 | 3,290 | +20% | 0 | 0 | — |
case-04 | fail→pass | 16,908 | 16,841 | -0% | 1 | 1 | 0% | 2,705 | 3,319 | +23% | 0 | 0 | — |
case-05 | fail→pass | 15,205 | 13,065 | -14% | 1 | 1 | 0% | 2,516 | 2,462 | -2% | 0 | 0 | — |
case-06 | fail→pass | 19,920 | 12,377 | -38% | 1 | 1 | 0% | 2,942 | 2,308 | -22% | 0 | 0 | — |
case-07 | fail→pass | 14,273 | 11,950 | -16% | 1 | 1 | 0% | 2,106 | 2,147 | +2% | 0 | 0 | — |
case-08 | fail→fail | 16,873 | 16,653 | -1% | 1 | 1 | 0% | 2,741 | 2,980 | +9% | 0 | 0 | — |
case-09 | pass→pass | 13,812 | 11,640 | -16% | 1 | 1 | 0% | 2,198 | 2,201 | +0% | 0 | 0 | — |
case-10 | fail→fail | 16,561 | 13,647 | -18% | 1 | 1 | 0% | 2,565 | 2,629 | +2% | 0 | 0 | — |
case-11 | pass→pass | 14,119 | 11,493 | -19% | 1 | 1 | 0% | 2,471 | 2,294 | -7% | 0 | 0 | — |
case-12 | pass→pass | 9,196 | 8,291 | -10% | 1 | 1 | 0% | 1,537 | 1,528 | -1% | 0 | 0 | — |
case-13 | pass→pass | 12,936 | 2,855 | -78% | 1 | 1 | 0% | 2,052 | 818 | -60% | 0 | 0 | — |
case-14 | pass→pass | 7,177 | 3,274 | -54% | 1 | 1 | 0% | 1,223 | 805 | -34% | 0 | 0 | — |
case-15 | pass→pass | 12,518 | 7,413 | -41% | 1 | 1 | 0% | 1,945 | 1,580 | -19% | 0 | 0 | — |
case-16 | pass→pass | 15,608 | 13,850 | -11% | 1 | 1 | 0% | 2,910 | 2,774 | -5% | 0 | 0 | — |
case-17 | pass→pass | 4,034 | 4,797 | +19% | 1 | 1 | 0% | 582 | 1,167 | +101% | 0 | 0 | — |
case-18 | pass→pass | 16,193 | 15,264 | -6% | 1 | 1 | 0% | 2,396 | 3,020 | +26% | 0 | 0 | — |
case-19 | pass→pass | 15,146 | 15,257 | +1% | 1 | 1 | 0% | 2,552 | 2,883 | +13% | 0 | 0 | — |
case-20 | pass→pass | 10,928 | 8,398 | -23% | 1 | 1 | 0% | 1,872 | 1,827 | -2% | 0 | 0 | — |
case-21 | pass→pass | 15,545 | 17,647 | +14% | 1 | 1 | 0% | 2,404 | 3,131 | +30% | 0 | 0 | — |
case-22 | fail→fail | 16,861 | 19,702 | +17% | 1 | 1 | 0% | 2,931 | 4,046 | +38% | 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. 22 cases were attempted. The headline lift of +32 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.