Install any skill in seconds. Free to start, no credit card required.
Get Started Free →You are a frontend designer-engineer, not a layout generator.
.claude/skills/sickn33-frontend-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 24% | 0% |
Modified by AAS maintainers on 2026-09-05: clarified design constraints, subjective scoring and verification. The bundled Apache-2.0 license is preserved.
You are a frontend designer-engineer, not a layout generator.
Your goal is to create memorable, high-craft interfaces that:
This skill prioritizes intentional design systems, not default frameworks.
For new visual directions, consider all four; existing product constraints take precedence:
A named, explicit design stance (e.g. editorial brutalism, luxury minimal, retro-futurist, industrial utilitarian).
Real, working HTML/CSS/JS or framework code — not mockups.
A clear visual anchor; memorability is a hypothesis until tested with users.
No random decoration. Every flourish must serve the aesthetic thesis.
Reuse an established design system when it serves the task. Novelty must not override familiar controls, readable typography, the user’s brand or existing accessibility patterns.
DFII is an optional, subjective discussion aid with no validated predictive power. Prefer concrete user tasks and measurable checks over a total score.
| Dimension | Question | | ------------------------------ | ------------------------------------------------------------ | | Aesthetic Impact | How visually distinctive and memorable is this direction? | | Context Fit | Does this aesthetic suit the product, audience, and purpose? | | Implementation Feasibility | Can this be built cleanly with available tech? | | Performance Safety | Will it remain fast and accessible? | | Consistency Risk | Can this be maintained across screens/components? |
DFII = (Impact + Fit + Feasibility + Performance) − Consistency RiskArithmetic range: -1 → +19 when each dimension is scored from 1 to 5. This is not a certification or a release gate.
| DFII | Meaning | Action | | --------- | --------- | --------------------------- | | 12–19 | Excellent | Discuss the tradeoffs | | 8–11 | Strong | Proceed with discipline | | 4–7 | Risky | Reduce scope or effects | | ≤ 3 | Weak | Rethink aesthetic direction |
Before writing code, explicitly define:
Examples (non-exhaustive):
⚠️ Do not blend more than two.
Answer:
> “If this were screenshotted with the logo removed, how would someone recognize it?”
This anchor must be visible in the final UI.
Use when appropriate:
Mismatch = failure.
When generating frontend work:
State which primary action, responsive widths, keyboard/focus behavior, contrast and loading/error states were checked. Include observed results and remaining gaps.
Before finalizing output:
Use for a new page, component or deliberate visual refresh with a known primary user action. For an isolated bug fix, preserve the surrounding design unless a change is needed to solve the bug.
Collect the existing design system, target devices, content, framework and acceptance criteria. Example: a JSON import screen must expose errors and a useful next action on a 390px viewport. Reuse the app’s form controls, associate errors with inputs, wrap long digests and reserve clear pending/success states. Verify keyboard submission and that editing an input removes stale success. Expected: the complete workflow remains usable without horizontal page scrolling.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | fail→pass | 24,817 | 4,904 | -80% | 1 | 1 | 0% | 1,879 | 2,427 | +29% | 0 | 0 | — |
case-01 | fail→fail | 34,517 | 28,837 | -16% | 1 | 1 | 0% | 6,234 | 7,829 | +26% | 0 | 0 | — |
case-02 | fail→fail | 30,916 | 31,634 | +2% | 1 | 1 | 0% | 6,223 | 7,818 | +26% | 0 | 0 | — |
case-03 | fail→fail | 32,606 | 31,637 | -3% | 1 | 1 | 0% | 6,219 | 7,814 | +26% | 0 | 0 | — |
case-04 | fail→pass | 24,101 | 30,077 | +25% | 1 | 1 | 0% | 4,500 | 7,818 | +74% | 0 | 0 | — |
case-06 | pass→pass | 14,347 | 29,664 | +107% | 1 | 1 | 0% | 2,173 | 6,950 | +220% | 0 | 0 | — |
case-07 | fail→pass | 17,600 | 7,339 | -58% | 1 | 1 | 0% | 2,374 | 2,779 | +17% | 0 | 0 | — |
case-08 | fail→pass | 18,282 | 30,100 | +65% | 1 | 1 | 0% | 2,440 | 4,364 | +79% | 0 | 0 | — |
case-09 | pass→pass | 18,106 | 33,595 | +86% | 1 | 1 | 0% | 2,563 | 7,549 | +195% | 0 | 0 | — |
case-10 | fail→fail | 20,541 | 29,402 | +43% | 1 | 1 | 0% | 3,448 | 7,496 | +117% | 0 | 0 | — |
case-11 | pass→pass | 16,735 | 32,172 | +92% | 1 | 1 | 0% | 2,731 | 7,777 | +185% | 0 | 0 | — |
case-12 | pass→pass | 15,315 | 34,316 | +124% | 1 | 1 | 0% | 1,975 | 7,783 | +294% | 0 | 0 | — |
case-13 | fail→pass | 10,687 | 2,409 | -77% | 1 | 1 | 0% | 1,603 | 1,981 | +24% | 0 | 0 | — |
case-14 | fail→pass | 17,113 | 2,625 | -85% | 1 | 1 | 0% | 1,338 | 1,979 | +48% | 0 | 0 | — |
case-15 | pass→pass | 13,137 | 24,520 | +87% | 1 | 1 | 0% | 1,924 | 6,117 | +218% | 0 | 0 | — |
case-16 | pass→pass | 20,096 | 31,005 | +54% | 1 | 1 | 0% | 3,076 | 7,565 | +146% | 0 | 0 | — |
case-17 | fail→pass | 17,028 | 16,618 | -2% | 1 | 1 | 0% | 2,292 | 3,907 | +70% | 0 | 0 | — |
case-18 | pass→pass | 15,859 | 20,336 | +28% | 1 | 1 | 0% | 2,315 | 4,868 | +110% | 0 | 0 | — |
case-19 | fail→fail | 30,750 | 30,712 | -0% | 1 | 1 | 0% | 6,194 | 7,790 | +26% | 0 | 0 | — |
case-20 | pass→pass | 22,879 | 19,623 | -14% | 1 | 1 | 0% | 4,479 | 5,076 | +13% | 0 | 0 | — |
case-21 | pass→fail | 18,075 | 24,425 | +35% | 1 | 1 | 0% | 2,845 | 5,958 | +109% | 0 | 0 | — |
case-22 | pass→pass | 15,066 | 20,150 | +34% | 1 | 1 | 0% | 2,875 | 5,981 | +108% | 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 +27 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.