Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate reusable UI patterns such as card sections, grids, lists, forms, and chart wrappers using StyleSeed Toss primitives.
.claude/skills/lingxling-ui-pattern/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -25% | 0% |
Part of StyleSeed, this skill builds reusable composed patterns from the seed's primitives. It is intended for sections like card lists, grids, form blocks, ranking lists, and chart wrappers that appear across multiple pages and need to look deliberate rather than ad hoc.
Common pattern families include:
Inspect both:
components/ui/ for primitivescomponents/patterns/ for neighboring patterns that can be extendedThe goal is composition, not duplication.
Keep the Toss seed defaults intact:
Expose data through props instead of hardcoding content. If a pattern has multiple variants, keep the API explicit and small.
Avoid page-specific assumptions unless the user explicitly wants a one-off section. If the markup only works on one route, it probably belongs in a page component, not a shared pattern.
Provide:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 39,922 | 23,267 | -42% | 1 | 1 | 0% | 6,651 | 5,676 | -15% | 0 | 0 | — |
case-02 | fail→pass | 31,300 | 18,931 | -40% | 1 | 1 | 0% | 6,602 | 4,520 | -32% | 0 | 0 | — |
case-03 | fail→pass | 51,646 | 26,181 | -49% | 1 | 1 | 0% | 8,164 | 6,073 | -26% | 0 | 0 | — |
case-04 | fail→pass | 34,848 | 20,735 | -40% | 1 | 1 | 0% | 5,725 | 3,952 | -31% | 0 | 0 | — |
case-05 | fail→fail | 38,547 | 25,557 | -34% | 1 | 1 | 0% | 8,230 | 5,202 | -37% | 0 | 0 | — |
case-06 | fail→pass | 26,350 | 23,616 | -10% | 1 | 1 | 0% | 5,837 | 4,361 | -25% | 0 | 0 | — |
case-07 | fail→fail | 38,449 | 23,493 | -39% | 1 | 1 | 0% | 8,230 | 5,695 | -31% | 0 | 0 | — |
case-08 | fail→fail | 37,002 | 40,026 | +8% | 1 | 1 | 0% | 7,869 | 7,177 | -9% | 0 | 0 | — |
case-09 | fail→pass | 26,582 | 31,812 | +20% | 1 | 1 | 0% | 5,482 | 5,571 | +2% | 0 | 0 | — |
case-10 | fail→pass | 33,453 | 17,592 | -47% | 1 | 1 | 0% | 5,455 | 4,053 | -26% | 0 | 0 | — |
case-11 | fail→fail | 40,491 | 20,613 | -49% | 1 | 1 | 0% | 6,338 | 4,927 | -22% | 0 | 0 | — |
case-12 | fail→pass | 37,222 | 28,427 | -24% | 1 | 1 | 0% | 8,227 | 6,828 | -17% | 0 | 0 | — |
case-13 | fail→fail | 44,457 | 21,401 | -52% | 1 | 1 | 0% | 7,148 | 4,978 | -30% | 0 | 0 | — |
case-14 | fail→pass | 19,128 | 15,324 | -20% | 1 | 1 | 0% | 4,526 | 3,805 | -16% | 0 | 0 | — |
case-15 | fail→fail | 51,177 | 28,337 | -45% | 1 | 1 | 0% | 8,225 | 5,104 | -38% | 0 | 0 | — |
case-16 | fail→pass | 49,711 | 32,707 | -34% | 1 | 1 | 0% | 8,224 | 4,164 | -49% | 0 | 0 | — |
case-17 | fail→fail | 34,344 | 43,447 | +27% | 1 | 1 | 0% | 8,202 | 8,857 | +8% | 0 | 0 | — |
case-18 | pass→pass | 35,169 | 22,546 | -36% | 1 | 1 | 0% | 5,555 | 5,478 | -1% | 0 | 0 | — |
case-19 | fail→pass | 36,189 | 44,132 | +22% | 1 | 1 | 0% | 8,227 | 4,943 | -40% | 0 | 0 | — |
case-20 | fail→pass | 40,006 | 33,171 | -17% | 1 | 1 | 0% | 6,309 | 5,959 | -6% | 0 | 0 | — |
case-21 | pass→pass | 37,634 | 17,875 | -53% | 1 | 1 | 0% | 6,125 | 3,460 | -44% | 0 | 0 | — |
case-22 | pass→pass | 23,371 | 23,339 | -0% | 1 | 1 | 0% | 3,195 | 4,342 | +36% | 0 | 0 | — |
case-23 | pass→pass | 22,824 | 14,652 | -36% | 1 | 1 | 0% | 3,131 | 4,110 | +31% | 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 +52 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.