Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Scaffold a new mobile-first page using StyleSeed Toss layout patterns, section rhythm, and existing shell components.
.claude/skills/lingxling-ui-page/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 125% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 29% | 0% |
Part of StyleSeed, this skill scaffolds a complete page or screen using the Toss seed's mobile-first composition rules. It keeps page structure consistent by building on the existing shell, top bar, bottom navigation, and card rhythm instead of producing disconnected sections.
Read the current page scaffolding patterns first, especially:
Clarify:
Every screen should have one dominant purpose.
Lay out the page from highest importance to lowest:
Avoid repeating the same section type mechanically from top to bottom.
Default layout choices:
max-w-[430px]bg-backgroundpx-6space-y-6Use existing ui/ and patterns/ components wherever possible. New pages should primarily orchestrate existing building blocks, not recreate them.
Return:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 11,406 | 22,804 | +100% | 1 | 1 | 0% | 1,812 | 5,412 | +199% | 0 | 0 | — |
case-01 | fail→fail | 36,138 | 34,527 | -4% | 1 | 1 | 0% | 7,536 | 8,013 | +6% | 0 | 0 | — |
case-02 | fail→pass | 24,837 | 23,106 | -7% | 1 | 1 | 0% | 4,512 | 5,390 | +19% | 0 | 0 | — |
case-03 | fail→fail | 39,856 | 33,554 | -16% | 1 | 1 | 0% | 8,242 | 7,924 | -4% | 0 | 0 | — |
case-05 | fail→pass | 44,649 | 34,137 | -24% | 1 | 1 | 0% | 7,897 | 7,570 | -4% | 0 | 0 | — |
case-06 | fail→pass | 9,634 | 17,423 | +81% | 1 | 1 | 0% | 1,809 | 4,072 | +125% | 0 | 0 | — |
case-07 | fail→pass | 8,642 | 3,992 | -54% | 1 | 1 | 0% | 1,554 | 1,487 | -4% | 0 | 0 | — |
case-08 | fail→fail | 9,167 | 6,137 | -33% | 1 | 1 | 0% | 1,516 | 1,803 | +19% | 0 | 0 | — |
case-09 | pass→pass | 9,804 | 11,688 | +19% | 1 | 1 | 0% | 1,791 | 2,984 | +67% | 0 | 0 | — |
case-10 | pass→pass | 14,512 | 15,484 | +7% | 1 | 1 | 0% | 2,255 | 3,237 | +44% | 0 | 0 | — |
case-11 | pass→pass | 16,419 | 16,340 | -0% | 1 | 1 | 0% | 2,310 | 3,592 | +55% | 0 | 0 | — |
case-12 | fail→pass | 12,396 | 8,765 | -29% | 1 | 1 | 0% | 1,843 | 2,379 | +29% | 0 | 0 | — |
case-13 | pass→pass | 12,231 | 9,984 | -18% | 1 | 1 | 0% | 2,071 | 2,591 | +25% | 0 | 0 | — |
case-14 | fail→pass | 16,147 | 5,767 | -64% | 1 | 1 | 0% | 2,546 | 1,621 | -36% | 0 | 0 | — |
case-15 | pass→pass | 16,616 | 17,521 | +5% | 1 | 1 | 0% | 2,358 | 3,336 | +41% | 0 | 0 | — |
case-16 | pass→pass | 21,267 | 19,579 | -8% | 1 | 1 | 0% | 3,437 | 3,810 | +11% | 0 | 0 | — |
case-17 | fail→pass | 17,038 | 15,763 | -7% | 1 | 1 | 0% | 2,666 | 3,504 | +31% | 0 | 0 | — |
case-18 | fail→fail | 8,225 | 13,071 | +59% | 1 | 1 | 0% | 1,324 | 3,104 | +134% | 0 | 0 | — |
case-19 | pass→pass | 12,567 | 7,844 | -38% | 1 | 1 | 0% | 1,773 | 1,937 | +9% | 0 | 0 | — |
case-20 | pass→pass | 17,402 | 19,114 | +10% | 1 | 1 | 0% | 3,302 | 4,759 | +44% | 0 | 0 | — |
case-21 | pass→pass | 12,724 | 10,298 | -19% | 1 | 1 | 0% | 2,427 | 2,573 | +6% | 0 | 0 | — |
case-22 | pass→pass | 6,261 | 7,137 | +14% | 1 | 1 | 0% | 1,201 | 2,138 | +78% | 0 | 0 | — |
case-23 | pass→pass | 7,451 | 7,821 | +5% | 1 | 1 | 0% | 1,394 | 2,269 | +63% | 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 +30 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.