Install any skill in seconds. Free to start, no credit card required.
Get Started Free →UI component and styling guidelines using Shadcn UI, Radix UI, and Tailwind
.claude/skills/elie222-ui-components/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 18% | 0% |
next/image package for imagesshpnpm dlx shadcn@latest add COMPONENT
Example:
shpnpm dlx shadcn@latest add progress
For API get requests to server use the swr package:
typescriptconst searchParams = useSearchParams(); const page = searchParams.get("page") || "1"; const { data, isLoading, error } = useSWR<PlanHistoryResponse>( `/api/user/planned/history?page=${page}` );
Use the LoadingContent component to handle loading states:
tsx<Card> <LoadingContent loading={isLoading} error={error}> {data && <MyComponent data={data} />} </LoadingContent> </Card>
tsx<Input type="email" name="email" label="Email" registerProps={register("email", { required: true })} error={errors.email} />
tsx<Input type="text" autosizeTextarea rows={3} name="message" placeholder="Paste in email content" registerProps={register("message", { required: true })} error={errors.message} />
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 22,111 | 15,310 | -31% | 1 | 1 | 0% | 4,566 | 3,967 | -13% | 0 | 0 | — |
case-02 | fail→pass | 13,587 | 21,995 | +62% | 1 | 1 | 0% | 2,984 | 5,191 | +74% | 0 | 0 | — |
case-03 | pass→pass | 12,988 | 5,487 | -58% | 1 | 1 | 0% | 2,454 | 1,499 | -39% | 0 | 0 | — |
case-04 | pass→pass | 11,120 | 4,651 | -58% | 1 | 1 | 0% | 2,124 | 1,252 | -41% | 0 | 0 | — |
case-05 | fail→pass | 15,137 | 10,132 | -33% | 1 | 1 | 0% | 2,899 | 2,445 | -16% | 0 | 0 | — |
case-06 | pass→pass | 6,942 | 2,747 | -60% | 1 | 1 | 0% | 1,154 | 721 | -38% | 0 | 0 | — |
case-07 | fail→pass | 8,916 | 3,741 | -58% | 1 | 1 | 0% | 1,514 | 1,033 | -32% | 0 | 0 | — |
case-08 | fail→pass | 17,349 | 16,309 | -6% | 1 | 1 | 0% | 3,193 | 3,761 | +18% | 0 | 0 | — |
case-09 | pass→pass | 20,616 | 19,652 | -5% | 1 | 1 | 0% | 3,459 | 3,378 | -2% | 0 | 0 | — |
case-10 | pass→pass | 18,124 | 3,082 | -83% | 1 | 1 | 0% | 3,264 | 869 | -73% | 0 | 0 | — |
case-11 | pass→pass | 2,781 | 1,980 | -29% | 1 | 1 | 0% | 465 | 628 | +35% | 0 | 0 | — |
case-12 | fail→pass | 18,175 | 14,603 | -20% | 1 | 1 | 0% | 2,881 | 2,875 | -0% | 0 | 0 | — |
case-13 | fail→pass | 16,202 | 17,567 | +8% | 1 | 1 | 0% | 3,366 | 3,245 | -4% | 0 | 0 | — |
case-14 | fail→fail | 15,702 | 20,804 | +32% | 1 | 1 | 0% | 3,002 | 4,880 | +63% | 0 | 0 | — |
case-15 | pass→pass | 16,166 | 6,166 | -62% | 1 | 1 | 0% | 2,745 | 1,435 | -48% | 0 | 0 | — |
case-16 | pass→pass | 5,034 | 1,345 | -73% | 1 | 1 | 0% | 800 | 541 | -32% | 0 | 0 | — |
case-17 | fail→pass | 11,552 | 4,409 | -62% | 1 | 1 | 0% | 2,195 | 1,258 | -43% | 0 | 0 | — |
case-18 | pass→pass | 12,182 | 8,250 | -32% | 1 | 1 | 0% | 2,215 | 2,031 | -8% | 0 | 0 | — |
case-19 | pass→pass | 14,990 | 9,926 | -34% | 1 | 1 | 0% | 2,299 | 2,049 | -11% | 0 | 0 | — |
case-20 | pass→pass | 11,238 | 12,752 | +13% | 1 | 1 | 0% | 2,166 | 2,388 | +10% | 0 | 0 | — |
case-21 | pass→pass | 14,210 | 13,813 | -3% | 1 | 1 | 0% | 2,596 | 2,944 | +13% | 0 | 0 | — |
case-22 | pass→pass | 11,724 | 8,107 | -31% | 1 | 1 | 0% | 2,175 | 1,846 | -15% | 0 | 0 | — |
case-23 | fail→pass | 15,435 | 7,497 | -51% | 1 | 1 | 0% | 2,910 | 1,799 | -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. 23 cases were attempted. The headline lift of +39 percentage points is the difference between those two pass rates over the 23 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.