Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Apply production-grade design standards when building or reviewing pages, components, or UI. Use this skill whenever the user asks to build a page, design a component, lay out a section, review the UI, fix the layout, or check design quality. Triggers on build a page, create a component, design a section, hero, card, CTA, layout, review the UI, fix the design, design system, design tokens, spacing, typography scale, button standards, mobile design. Also triggers for any production design decisio
.claude/skills/rampstackco-design-standards/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 66% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 108% | 0% |
| case-10 | ✓→✓ | = Same ✓ | 83% | 0% |
Production design standards for any web project. Stack-agnostic. Tool-agnostic. Covers tokens, contrast, hierarchy, spacing, mobile rules, and the pre-ship checklist.
This skill complements brand-identity (which defines the visual system) and brand-style-guide (which documents it). This skill is for the day-to-day work of applying those standards in real components and pages.
brand-identity)brand-style-guide)design-system)frontend-component-build)accessibility-audit)If brand tokens are undefined, define a working set first using the template in references/design-tokens-template.md.
Six standards apply to every piece of UI. Hold the line on these and most design problems disappear.
Every project needs tokens defined before any UI gets built. Tokens are the source of truth.
Color tokens (minimum):
Spacing tokens:
Type tokens:
Radius tokens:
Document the tokens once. Reference them everywhere. Hardcoded values are technical debt.
Contrast is not a preference. It is a baseline. A design that fails AA is broken, regardless of how it looks to designers with full vision.
| Element | Required ratio | |---|---| | Normal body text | 4.5:1 | | Large text (18pt regular or 14pt bold) | 3:1 | | UI components and graphical elements | 3:1 |
Common failures to avoid:
For the math, the contrast checker, and brand-color strategies, see brand-identity/references/contrast-and-accessibility.md.
A well-designed page has a clear scan order. Hierarchy comes from:
Apply hierarchy intentionally. Every visual element should be reachable through the hierarchy. If three things compete for primary attention, none of them wins.
Common failures:
Spacing is what separates a polished layout from a cluttered one.
Rules:
Common failures:
Most users are on mobile. Designing for desktop first guarantees mobile failures.
Standards:
Common failures:
The same thing should look the same everywhere. Variations should be intentional.
Common consistency principles:
Common failures:
references/preship-checklist.md.This skill produces design decisions and review notes more than artifacts. Outputs include:
design-tokens.md or equivalentWhen generating component code (HTML, CSS, framework-specific markup), the SKILL.md remains stack-agnostic. Stack-specific patterns live in reference files.
references/design-tokens-template.md - Template for setting up tokens for any project.references/preship-checklist.md - Final design review checklist before shipping.references/tailwind-patterns.md - Optional. Tailwind-specific component patterns (hero, cards, buttons, data rows) for projects on that stack.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 15,979 | 20,876 | +31% | 1 | 1 | 0% | 2,495 | 5,609 | +125% | 0 | 0 | — |
case-02 | fail→fail | 23,992 | 25,267 | +5% | 1 | 1 | 0% | 4,706 | 7,031 | +49% | 0 | 0 | — |
case-03 | fail→fail | 21,855 | 23,570 | +8% | 1 | 1 | 0% | 3,947 | 6,334 | +60% | 0 | 0 | — |
case-04 | pass→pass | 14,247 | 10,797 | -24% | 1 | 1 | 0% | 2,335 | 3,868 | +66% | 0 | 0 | — |
case-05 | pass→pass | 11,791 | 11,110 | -6% | 1 | 1 | 0% | 1,903 | 3,956 | +108% | 0 | 0 | — |
case-10 | pass→pass | 8,379 | 2,235 | -73% | 1 | 1 | 0% | 1,339 | 2,452 | +83% | 0 | 0 | — |
case-06 | pass→pass | 9,047 | 8,027 | -11% | 1 | 1 | 0% | 1,503 | 3,360 | +124% | 0 | 0 | — |
case-07 | pass→pass | 9,113 | 6,938 | -24% | 1 | 1 | 0% | 1,487 | 3,290 | +121% | 0 | 0 | — |
case-08 | pass→pass | 10,432 | 7,728 | -26% | 1 | 1 | 0% | 1,707 | 3,215 | +88% | 0 | 0 | — |
case-09 | pass→pass | 12,274 | 11,977 | -2% | 1 | 1 | 0% | 1,735 | 3,950 | +128% | 0 | 0 | — |
case-11 | pass→pass | 7,657 | 5,937 | -22% | 1 | 1 | 0% | 1,347 | 3,123 | +132% | 0 | 0 | — |
case-12 | fail→pass | 14,561 | 7,530 | -48% | 1 | 1 | 0% | 2,245 | 3,274 | +46% | 0 | 0 | — |
case-13 | fail→fail | 12,901 | 10,646 | -17% | 1 | 1 | 0% | 2,003 | 3,836 | +92% | 0 | 0 | — |
case-14 | pass→pass | 10,122 | 7,960 | -21% | 1 | 1 | 0% | 1,654 | 3,444 | +108% | 0 | 0 | — |
case-15 | pass→pass | 11,337 | 9,525 | -16% | 1 | 1 | 0% | 1,644 | 3,660 | +123% | 0 | 0 | — |
case-16 | fail→fail | 19,190 | 12,010 | -37% | 1 | 1 | 0% | 3,078 | 4,169 | +35% | 0 | 0 | — |
case-17 | fail→fail | 16,391 | 13,613 | -17% | 1 | 1 | 0% | 2,897 | 4,540 | +57% | 0 | 0 | — |
case-18 | fail→pass | 13,126 | 4,844 | -63% | 1 | 1 | 0% | 2,032 | 2,915 | +43% | 0 | 0 | — |
case-19 | pass→pass | 13,036 | 13,161 | +1% | 1 | 1 | 0% | 2,098 | 4,196 | +100% | 0 | 0 | — |
case-20 | pass→pass | 13,380 | 13,164 | -2% | 1 | 1 | 0% | 2,077 | 4,184 | +101% | 0 | 0 | — |
case-21 | pass→pass | 12,593 | 9,446 | -25% | 1 | 1 | 0% | 1,967 | 3,678 | +87% | 0 | 0 | — |
case-22 | pass→pass | 9,558 | 4,681 | -51% | 1 | 1 | 0% | 1,452 | 2,853 | +96% | 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 +9 percentage points is the difference between those two pass rates over the 22 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.