Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.
.claude/skills/majiayu000-frontend-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 16% | 0% |
This skill guides creation of distinctive, production-grade frontend interfaces that avoid generic "AI slop" aesthetics. Implement real working code with exceptional attention to aesthetic details and creative choices.
The user provides frontend requirements: a component, page, application, or interface to build. They may include context about the purpose, audience, or technical constraints.
Use this skill only when the implementation target is a web frontend. Before applying its aesthetics guidance, verify that the relevant code is HTML/CSS/JavaScript or a web UI framework.
Stop using this skill if repository inspection shows that the visual surface is implemented by a native renderer, GPU shader, game engine, terminal UI, video pipeline, or another non-web graphics stack. Route that work to the relevant language, rendering, or visual-verification workflow instead. A request to make something look more polished does not establish web-frontend scope.
Trigger regression cases live in evals/triggers.jsonl.
Before coding, understand the context and commit to a BOLD aesthetic direction:
CRITICAL: Choose a clear conceptual direction and execute it with precision. Bold maximalism and refined minimalism both work - the key is intentionality, not intensity.
Then implement working code (HTML/CSS/JS, React, Vue, etc.) that is:
Focus on:
NEVER use generic AI-generated aesthetics like overused font families (Inter, Roboto, Arial, system fonts), cliched color schemes (particularly purple gradients on white backgrounds), predictable layouts and component patterns, and cookie-cutter design that lacks context-specific character.
Interpret creatively and make unexpected choices that feel genuinely designed for the context. No design should be the same. Vary between light and dark themes, different fonts, different aesthetics. NEVER converge on common choices (Space Grotesk, for example) across generations.
IMPORTANT: Match implementation complexity to the aesthetic vision. Maximalist designs need elaborate code with extensive animations and effects. Minimalist or refined designs need restraint, precision, and careful attention to spacing, typography, and subtle details. Elegance comes from executing the vision well.
Remember: Claude is capable of extraordinary creative work. Don't hold back, show what can truly be created when thinking outside the box and committing fully to a distinctive vision.
Before completion, check the implemented interface in the target runtime whenever possible. Verify layout, responsive behavior, text fit, contrast, interaction states, and that the result is a working UI rather than a static mockup. For local apps, use the available browser or screenshot workflow to inspect desktop and mobile viewports before claiming the frontend is done.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 31,211 | 31,942 | +2% | 1 | 1 | 0% | 6,191 | 7,173 | +16% | 0 | 0 | — |
case-01 | fail→fail | 33,275 | 37,812 | +14% | 1 | 1 | 0% | 6,201 | 7,183 | +16% | 0 | 0 | — |
case-02 | fail→pass | 31,297 | 29,537 | -6% | 1 | 1 | 0% | 6,197 | 7,179 | +16% | 0 | 0 | — |
case-03 | fail→pass | 30,365 | 32,859 | +8% | 1 | 1 | 0% | 6,194 | 7,176 | +16% | 0 | 0 | — |
case-05 | fail→pass | 30,361 | 31,494 | +4% | 1 | 1 | 0% | 6,175 | 7,157 | +16% | 0 | 0 | — |
case-06 | pass→pass | 28,586 | 34,015 | +19% | 1 | 1 | 0% | 6,171 | 7,153 | +16% | 0 | 0 | — |
case-07 | fail→fail | 31,985 | 31,598 | -1% | 1 | 1 | 0% | 6,173 | 7,156 | +16% | 0 | 0 | — |
case-08 | fail→fail | 34,748 | 32,662 | -6% | 1 | 1 | 0% | 6,169 | 7,151 | +16% | 0 | 0 | — |
case-09 | pass→pass | 30,298 | 32,525 | +7% | 1 | 1 | 0% | 6,172 | 7,154 | +16% | 0 | 0 | — |
case-10 | pass→pass | 31,631 | 33,392 | +6% | 1 | 1 | 0% | 6,168 | 7,150 | +16% | 0 | 0 | — |
case-11 | pass→pass | 31,307 | 32,492 | +4% | 1 | 1 | 0% | 6,168 | 7,150 | +16% | 0 | 0 | — |
case-12 | pass→pass | 35,872 | 31,549 | -12% | 1 | 1 | 0% | 6,174 | 7,156 | +16% | 0 | 0 | — |
case-13 | fail→pass | 29,362 | 31,597 | +8% | 1 | 1 | 0% | 6,165 | 7,147 | +16% | 0 | 0 | — |
case-14 | fail→pass | 28,738 | 29,879 | +4% | 1 | 1 | 0% | 6,166 | 7,148 | +16% | 0 | 0 | — |
case-15 | fail→fail | 32,365 | 31,278 | -3% | 1 | 1 | 0% | 6,162 | 7,144 | +16% | 0 | 0 | — |
case-16 | pass→fail | 28,438 | 31,137 | +9% | 1 | 1 | 0% | 6,157 | 7,139 | +16% | 0 | 0 | — |
case-17 | fail→pass | 28,005 | 31,714 | +13% | 1 | 1 | 0% | 6,159 | 7,141 | +16% | 0 | 0 | — |
case-18 | pass→pass | 23,675 | 29,484 | +25% | 1 | 1 | 0% | 5,035 | 7,144 | +42% | 0 | 0 | — |
case-19 | fail→fail | 30,314 | 31,707 | +5% | 1 | 1 | 0% | 6,161 | 7,143 | +16% | 0 | 0 | — |
case-20 | pass→pass | 27,529 | 23,008 | -16% | 1 | 1 | 0% | 5,553 | 6,272 | +13% | 0 | 0 | — |
case-21 | pass→fail | 27,891 | 32,669 | +17% | 1 | 1 | 0% | 6,014 | 7,151 | +19% | 0 | 0 | — |
case-22 | pass→pass | 35,261 | 29,176 | -17% | 1 | 1 | 0% | 5,594 | 7,148 | +28% | 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 +18 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 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.