Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Implement UI using Shadcn MCP (atoms/theme) and 21st.dev MCP (complex sections). Use when adding buttons, layouts, or generating landing pages.
.claude/skills/aiskillstore-ui-handler/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-23 | ✓→✗ | ▼ Worse | -82% | 0% |
Use the Shadcn MCP for foundational UI elements (buttons, inputs, dialogs).
> "Add the button and dialog components using Shadcn MCP."
pnpm dlx shadcn@latest add {component}Use the 21st.dev Magic MCP for high-level sections (Landing pages, Heros, Dashboards).
> "Generate a modern SaaS hero section with a dark gradient background and email capture form."
src/components/sections/ or src/components/website/.page.tsx.Use Shadcn MCP or edit src/app/globals.css directly.
> "Update the app theme to use a 'zinc' neutral base with 'blue' primary color."
src/app/globals.css. Ensure any new styles use these variables (e.g., bg-background, text-primary).@/components/ui and sections from @/components/sections._components/ next to the page.For detailed architecture and best practices, see reference.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | pass→pass | 7,733 | 1,523 | -80% | 1 | 1 | 0% | 1,277 | 667 | -48% | 0 | 0 | — |
case-01 | fail→fail | 12,974 | 8,923 | -31% | 1 | 1 | 0% | 2,541 | 2,076 | -18% | 0 | 0 | — |
case-02 | fail→fail | 29,866 | 4,870 | -84% | 1 | 1 | 0% | 6,179 | 700 | -89% | 0 | 0 | — |
case-03 | fail→fail | 12,323 | 8,086 | -34% | 1 | 1 | 0% | 2,309 | 2,162 | -6% | 0 | 0 | — |
case-04 | fail→fail | 15,637 | 7,466 | -52% | 1 | 1 | 0% | 2,853 | 1,937 | -32% | 0 | 0 | — |
case-05 | fail→pass | 14,885 | 4,456 | -70% | 1 | 1 | 0% | 2,296 | 1,139 | -50% | 0 | 0 | — |
case-06 | fail→pass | 42,909 | 13,734 | -68% | 1 | 1 | 0% | 3,893 | 3,337 | -14% | 0 | 0 | — |
case-07 | pass→pass | 8,793 | 3,515 | -60% | 1 | 1 | 0% | 1,526 | 1,005 | -34% | 0 | 0 | — |
case-18 | pass→pass | 10,040 | 2,603 | -74% | 1 | 1 | 0% | 1,708 | 889 | -48% | 0 | 0 | — |
case-08 | pass→pass | 7,543 | 1,979 | -74% | 1 | 1 | 0% | 1,296 | 753 | -42% | 0 | 0 | — |
case-09 | pass→pass | 3,773 | 1,477 | -61% | 1 | 1 | 0% | 615 | 668 | +9% | 0 | 0 | — |
case-10 | pass→pass | 14,300 | 2,137 | -85% | 1 | 1 | 0% | 2,304 | 771 | -67% | 0 | 0 | — |
case-11 | pass→pass | 5,651 | 1,954 | -65% | 1 | 1 | 0% | 1,073 | 736 | -31% | 0 | 0 | — |
case-19 | fail→pass | 11,728 | 8,312 | -29% | 1 | 1 | 0% | 1,773 | 1,716 | -3% | 0 | 0 | — |
case-12 | fail→pass | 10,125 | 2,919 | -71% | 1 | 1 | 0% | 1,587 | 879 | -45% | 0 | 0 | — |
case-13 | fail→fail | 19,025 | 11,503 | -40% | 1 | 1 | 0% | 3,904 | 3,047 | -22% | 0 | 0 | — |
case-14 | pass→pass | 7,289 | 2,825 | -61% | 1 | 1 | 0% | 1,163 | 827 | -29% | 0 | 0 | — |
case-15 | pass→pass | 2,725 | 1,564 | -43% | 1 | 1 | 0% | 471 | 693 | +47% | 0 | 0 | — |
case-16 | pass→pass | 6,835 | 3,277 | -52% | 1 | 1 | 0% | 1,162 | 990 | -15% | 0 | 0 | — |
case-20 | pass→pass | 14,947 | 9,958 | -33% | 1 | 1 | 0% | 2,871 | 2,460 | -14% | 0 | 0 | — |
case-21 | pass→pass | 9,653 | 7,609 | -21% | 1 | 1 | 0% | 1,700 | 1,866 | +10% | 0 | 0 | — |
case-22 | pass→pass | 17,380 | 19,241 | +11% | 1 | 1 | 0% | 3,492 | 3,476 | -0% | 0 | 0 | — |
case-23 | pass→fail | 17,130 | 3,297 | -81% | 1 | 1 | 0% | 3,555 | 633 | -82% | 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, and 21 counted toward the lift figure. The other 2 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 +13 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is 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.