Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Produce frontend UI that actually looks designed — a working spacing/type system, deliberate color use, real states, and restraint — instead of the generic AI-generated interface. Use when asked to build or restyle a UI, landing page, dashboard, or component, when output 'works but looks like a prototype', or to establish the visual system for a new app. Produces working HTML/CSS (or framework components) built on an explicit token system, with hover/focus/empty/loading states included. For crit
.claude/skills/mohitagw15856-frontend-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 575% | 0% |
| case-05 | ✓→✗ | ▼ Worse | 130% | 0% |
| case-06 | ✓→✗ | ▼ Worse | 405% | 0% |
AI-generated UI has a recognisable smell: default blues, five different paddings, everything the same visual weight, no states. This skill produces interfaces that look decided — by making the decisions explicit as a token system, then spending contrast deliberately instead of everywhere.
Ask for (if not already provided):
4/8/12/16/24/32/48/64. Every margin/padding/gap comes FROM the scale — the #1 tell of undesigned UI is seventeen distinct paddings. Related things sit closer than unrelated things (proximity is free information design).bg / surface / border / text / text-muted / accent / danger / success. ONE accent, spent where attention belongs — the primary action, the active state, the number that matters. The 90% of a designed UI is neutrals; if everything is colorful, nothing is. Check text contrast (4.5:1 body, 3:1 large) as you pick, not after.prefers-reduced-motion respected; nothing bounces in a pro tool.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 25,942 | 26,484 | +2% | 1 | 1 | 0% | 6,220 | 7,305 | +17% | 0 | 0 | — |
case-02 | fail→fail | 25,105 | 25,496 | +2% | 1 | 1 | 0% | 6,214 | 7,298 | +17% | 0 | 0 | — |
case-03 | fail→fail | 24,425 | 26,142 | +7% | 1 | 1 | 0% | 6,204 | 7,289 | +17% | 0 | 0 | — |
case-04 | pass→fail | 5,478 | 24,678 | +350% | 1 | 1 | 0% | 1,072 | 7,240 | +575% | 0 | 0 | — |
case-05 | pass→fail | 6,556 | 8,263 | +26% | 1 | 1 | 0% | 1,307 | 3,002 | +130% | 0 | 0 | — |
case-06 | pass→fail | 7,359 | 25,577 | +248% | 1 | 1 | 0% | 1,434 | 7,241 | +405% | 0 | 0 | — |
case-07 | fail→pass | 25,093 | 23,956 | -5% | 1 | 1 | 0% | 6,180 | 7,265 | +18% | 0 | 0 | — |
case-08 | fail→pass | 25,321 | 26,784 | +6% | 1 | 1 | 0% | 6,158 | 7,256 | +18% | 0 | 0 | — |
case-09 | fail→fail | 26,691 | 25,333 | -5% | 1 | 1 | 0% | 6,171 | 7,256 | +18% | 0 | 0 | — |
case-10 | fail→fail | 22,338 | 24,357 | +9% | 1 | 1 | 0% | 5,371 | 7,256 | +35% | 0 | 0 | — |
case-11 | fail→fail | 25,564 | 26,554 | +4% | 1 | 1 | 0% | 6,170 | 7,256 | +18% | 0 | 0 | — |
case-12 | fail→fail | 25,030 | 25,575 | +2% | 1 | 1 | 0% | 5,309 | 7,245 | +36% | 0 | 0 | — |
case-13 | fail→fail | 26,747 | 25,556 | -4% | 1 | 1 | 0% | 6,163 | 7,248 | +18% | 0 | 0 | — |
case-14 | pass→fail | 20,399 | 27,307 | +34% | 1 | 1 | 0% | 4,411 | 7,254 | +64% | 0 | 0 | — |
case-15 | fail→fail | 26,559 | 27,340 | +3% | 1 | 1 | 0% | 5,835 | 7,249 | +24% | 0 | 0 | — |
case-22 | pass→pass | 14,537 | 25,011 | +72% | 1 | 1 | 0% | 3,197 | 7,245 | +127% | 0 | 0 | — |
case-16 | pass→pass | 18,142 | 27,263 | +50% | 1 | 1 | 0% | 4,526 | 7,251 | +60% | 0 | 0 | — |
case-17 | fail→fail | 24,922 | 26,335 | +6% | 1 | 1 | 0% | 6,160 | 7,245 | +18% | 0 | 0 | — |
case-18 | fail→fail | 16,775 | 27,089 | +61% | 1 | 1 | 0% | 3,243 | 7,255 | +124% | 0 | 0 | — |
case-19 | pass→fail | 23,194 | 25,533 | +10% | 1 | 1 | 0% | 5,453 | 7,241 | +33% | 0 | 0 | — |
case-20 | pass→fail | 24,945 | 24,933 | -0% | 1 | 1 | 0% | 6,154 | 7,239 | +18% | 0 | 0 | — |
case-21 | fail→fail | 28,857 | 27,316 | -5% | 1 | 1 | 0% | 6,159 | 7,244 | +18% | 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 -33 percentage points is the difference between those two pass rates over the 22 comparable cases. 8 cases got worse with the skill loaded, and they are 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.