agentleFS
SirCharanSirCharan/claude-code-harness/skills/design-system/SKILL.md

design-system

Skill0 starsChanged 5 months agoMITView on GitHub

START HERE before building, scaffolding, or restyling ANY web UI / Vercel project (dashboard, marketing/landing microsite, editorial or research report, calculator/tool, docs site, or creative side-project). This is the master router + shared foundations for the design skills: it picks the right per-type skill, sets the house stack (Next.js + Tailwind v4 + shadcn/ui + Lucide, Recharts for data), kills system-ui defaults, and ships a paste-ready token layer, WCAG 2.2 a11y baseline, component baseline, motion + responsive defaults, a named-style picker, and a "premium-feel" checklist distilled from Stripe/Vercel/Linear/Robinhood/Apple/Mercury. Trigger on: "build a UI", "new landing page / dashboard / report / calculator / docs site", "new Vercel project", "design system", "make this look better", "this looks templated", "pick a palette/font/style", "upgrade this ChatGPT/Grok project", or any web design work.

192 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use setup-engine, team-level and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • CoralGameSkill

    setup-engine

    Configure the project's game engine and version. Pins the engine in CLAUDE.md, detects knowledge gaps, and populates engine reference docs via WebSearch when the version is beyond the LLM's training data.

    Also in 13 projects that use this

  • CoralGameSkill

    team-level

    Orchestrate level design team: level-designer + narrative-director + world-builder + art-director + systems-designer + qa-tester for complete area/level creation.

    Also in 12 projects that use this

  • lilangMaxSkill

    create-stories

    Break a single epic into implementable story files. Reads the epic, its GDD, governing ADRs, and control manifest. Each story embeds its GDD requirement TR-ID, ADR guidance, acceptance criteria, story type, and test evidence path. Run after /create-epics for each epic.

    Also in 11 projects that use this

  • lilangMaxSkill

    story-done

    End-of-story completion review. Reads the story file, verifies each acceptance criterion against the implementation, checks for GDD/ADR deviations, prompts code review, updates story status to Complete, and surfaces the next ready story from the sprint.

    Also in 11 projects that use this

  • +2 more from CoralGame
  • +6 more from lilangMax

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 175 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If two could apply (e.g. a dashboard that's also a marketing page), load both and let the dominant surface win.

    line 46
  2. Distinctive:

    Verify live on Vercel + run the §5 a11y checklist before calling it done.

    line 212
  3. Distinctive:

    Treat ◐ as strong defaults.

    line 220

“Common” is measured: kept by at least 6 in 10 of the 175 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Design System — master router + shared foundations

Researched 2026 baseline for every web UI. Confidence per claim: ✅ = verified from a primary/teardown source; ◐ = well-sourced design judgment (strong default, not dogma).

Always do this first: (1) classify the project type → load the matching skill below; (2) load layout-interaction-design for any non-trivial interactive/multi-screen build (spacing, button placement, navigation, states); (3) apply the shared foundations + premium-feel checklist from THIS skill on top. Division of labor: the per-type skill owns its layout/UX; layout-interaction-design owns spacing mechanics, action hierarchy, routing & states; THIS skill owns tokens, type, a11y, motion, components, and the overall look.


1. Router — pick the project type

Building…Load this skillPlus
Data/analytics dashboard, trading/signal UI, chartsdataviz-designthis skill
Marketing / landing / GTM micrositemarketing-microsite-designthis skill
Editorial / research report, almanac, explainer, long-formeditorial-report-designthis skill
Calculator / single-purpose tooltool-calculator-designthis skill
Docs site (Fumadocs/Nextra)docs-designthis skill
Creative / culture side-project— go expressive; §6 "bold" stylesthis skill (a11y only)

For EVERY type (the behavioral layer): also load layout-interaction-design — spacing/padding/ margin/gap mechanics, button & action placement (incl. the cross-platform order disagreement), navigation & routing (Next.js App Router: Link/prefetch, layouts, loading.tsx, modal routes, active link), and the full interaction-state set (loading/empty/error, focus management, feedback latency).

If two could apply (e.g. a dashboard that's also a marketing page), load both and let the dominant surface win.


2. House stack & non-negotiables ✅◐

  • Stack: Next.js (App Router) + Tailwind v4 (CSS-first, no tailwind.config.js) + shadcn/ui (init with Base UI primitives in 2026: npx shadcn@latest init --base-ui) + Lucide icons + Recharts for data. Motion via CSS first, motion (ex-Framer) only when needed (§5).
  • Never ship system-ui as the brand font. Pick a real font pair per type (see the per-type skill or §4). This single change separates the "product" tier from the "templated" tier.
  • Token-first. Define semantic CSS variables once; reference them everywhere; never hardcode hex in components or charts.
  • Dark vs light: dashboards & trading = dark-first; tools & docs = either; marketing & editorial = often light-first but support both. Never auto-invert semantic colors — re-map them.

3. Token layer — paste-ready (shadcn + Tailwind v4) ◐

Premium rules baked in: one saturated accent, everything else neutral; never pure #000/#fff for text/bg; 8px spacing grid; small consistent radii. Swap --accent per project; keep the rest.

/* globals.css */
@import "tailwindcss";

@theme inline {
  --color-bg: var(--bg);            --color-surface: var(--surface);
  --color-surface-2: var(--surface-2); --color-elevated: var(--elevated);
  --color-ink: var(--ink);          --color-ink-2: var(--ink-2);
  --color-ink-muted: var(--ink-muted); --color-border: var(--border);
  --color-accent: var(--accent);    --color-accent-hover: var(--accent-hover);
  --color-ring: var(--accent);
  --color-pos: var(--pos); --color-neg: var(--neg); --color-warn: var(--warn);
  --radius-sm: 4px; --radius-md: 6px; --radius-lg: 8px; --radius-pill: 9999px;
  --ease-out: cubic-bezier(.16,1,.3,1); --ease-standard: cubic-bezier(.2,0,0,1);
  --dur-fast: 120ms; --dur-base: 200ms; --dur-slow: 320ms;
}

:root {                                  /* LIGHT — never pure white/black */
  --bg:#FFFFFF; --surface:#F6F8FA; --surface-2:#EDF0F3; --elevated:#FFFFFF;
  --ink:#0A1628; --ink-2:#3A4452; --ink-muted:#6B7280; --border:rgba(10,22,40,.10);
  --accent:#635BFF; --accent-hover:#5147E6;      /* ← change per project */
  --pos:#0E9F6E; --neg:#E02424; --warn:#C2700A;
}
.dark {                                  /* DARK — never pure black */
  --bg:#0D0D0F; --surface:#131317; --surface-2:#1A1A20; --elevated:#222229;
  --ink:#F0F0F5; --ink-2:#B0B0BF; --ink-muted:#84848F; --border:rgba(255,255,255,.08); /* ink-muted ≥4.5:1 on dark surfaces */
  --accent:#7C83FF; --accent-hover:#9AA0FF;
  --pos:#22C97A; --neg:#F04F5A; --warn:#E0A800;
}
  • Spacing: 8px grid → 4 · 8 · 12 · 16 · 24 · 32 · 48 · 64 · 96 · 128. Section padding ≥24–32px; let dense content sit inside roomy chrome.
  • Borders: hairlines, not <hr> — 1px at low alpha (--border). Shadows: minimal; none on flat marketing surfaces; soft layered shadow only for true elevation (cards/popovers/modals).
  • Radii: small + consistent (4/6/8). Reserve pill for marketing CTAs, not product chrome.
  • For dense dark dashboards, dataviz-design §2 has the deeper surface/ink ladder, CVD-safe diverging P&L, and heatmap ramps — use those there.

4. Typography ◐ (full detail: per-type skill + research)

  • One modular ratio: UI/dashboards 1.2; marketing/editorial 1.25–1.333. Use clamp() fluid type, keep a rem term so browser zoom works, body ≥16px.
  • Measure: body max-width: 60–70ch (66 ideal); line-height 1.5–1.65 body, 1.05–1.2 headings.
  • Tracking: negative on large headings only (-0.02 to -0.03em ≥24px); 0 body; +0.05em on caps.
  • Numerics: font-variant-numeric: tabular-nums on every figure (prices, %, P&L, dates) so digits don't jitter. Set on the container, not per cell.
  • Variable fonts + font-optical-sizing: auto on opsz families (Fraunces, Newsreader, Source Serif 4).
  • Default pairs (all via next/font/google): dashboards/tools → Geist + Geist Mono or Inter + IBM Plex Mono; marketing → Bricolage Grotesque / Fraunces display + Geist/Inter body; editorial → Newsreader / Source Serif 4 body + Inter UI; docs → Inter/Geist + JetBrains Mono.

5. A11y, components, motion, responsive ✅◐

A11y — WCAG 2.2 AA baseline (always):

  • Contrast: body ≥4.5:1, large/UI ≥3:1 (check both themes). Targets ≥24px (44px primary touch).
  • Visible :focus-visible ring; never bare outline:none. Logical tab order; one <h1>/page; landmarks.
  • Color is never the only signal — pair P&L color with ▲▼/sign. Honor prefers-reduced-motion globally.
  • Live updates (price/signal ticks) → aria-live="polite" / role="status".
:where(a,button,input,select,textarea,[tabindex]):focus-visible{
  outline:2px solid var(--accent); outline-offset:2px; border-radius:4px; }
@media (prefers-reduced-motion: reduce){
  *,*::before,*::after{ animation-duration:.01ms!important; animation-iteration-count:1!important;
    transition-duration:.01ms!important; scroll-behavior:auto!important; } }

Components: shadcn/ui (own the code; components/ui/* IS your design system). Compose with cva variants + cn(). Tailwind v4 gotchas: theme in @theme (no config.js), OKLCH tokens, [data-slot] styling hooks. Every interactive element ships 6 states: default, hover, focus, active, disabled, loading.

Motion: CSS transitions for the 90% (hover ~120ms, base ~200ms, slow ~320ms cap, ease-out). motion only for spring/gesture/layout/AnimatePresence exit orchestration. View Transitions API (Baseline Oct 2025) for route transitions. Asymmetric: enter ~instant/200ms, exit ~150ms opacity. Spatial/origin-aware (popovers scale from trigger). No transitions on long lists/tables.

Responsive: mobile-first viewport breakpoints for page layout; Tailwind v4 built-in container queries (@container) for component layout. Data-dense surfaces get a compact data-density mode (tighter spacing, ≤32px rows) but never below the 24px target floor.


6. Style picker — name the look, then prompt it ✅◐

Build the bones in a credible style, then add one texture accent in a hero/scorecard. Reserve bold styles for side-projects.

Fintech-safe core (use freely): Swiss/International (grid, sans, typography-as-hero — trust), Minimalism (whitespace, restraint — premium calm), Editorial/Magazine (serif display + pull quotes — authority), Bento grid (modular tiles — organized, "keynote"), Flat/Material (safe default), Glassmorphism ◐ (frosted cards — the premium surface accent, use lightly).

Use sparingly / side-projects only: ◐ Neumorphism, Claymorphism, Skeuomorphism, Neo-Brutalism, Brutalism, Maximalism, Aurora/mesh, Y2K, Synthwave, Memphis, Japandi.

TypePrimary styles
DashboardSwiss grid + dark Utilitarian; Glass/Bento for scorecards only
MarketingBento + Minimalism; Aurora/Glass hero accent
Editorial reportEditorial/Magazine + Swiss; Luxury serif for the cover
Calculator/toolMinimalism / Flat + Swiss; one Glass result card
DocsSwiss + Minimalism; Flat-Material components
Side-projectNeo-Brutalism / Y2K / Synthwave / Maximalism — distinctiveness > credibility

7. Premium-feel checklist — apply to every build ✅

Distilled from Stripe, Vercel/Geist, Linear, Robinhood, Apple, Mercury, Arc. If a generic Tailwind UI feels "templated," it's usually failing these:

  1. One saturated accent, everything else neutral — links/primary/active/focus only.
  2. Never pure black/white for text or bg — deep slate (#0A1628) / near-black (#0D0D0F); #000/#fff are only the extreme anchors.
  3. 8px spacing grid; generous section padding; dense content inside roomy chrome.
  4. 4–6 size type ramp, one family; negative tracking on large headings; tabular-nums on all figures.
  5. Semantic color = meaning (green=up, red=down) applied consistently — never decoration.
  6. Transitions 100–250ms ease-out, asymmetric (enter ~instant, exit ~150ms), spatial/origin-aware, none on long lists.
  7. 6 states per interactive element; visible keyboard focus.
  8. Hairline 1px borders at low alpha; minimal/zero shadows except true elevation.
  9. Small consistent radii (4/6/8); pill only for marketing CTAs.
  10. Whitespace = hierarchy — let content lead, chrome recede.
  11. Constrained gradients ◐ — ≤3 complementary stops, accents/hero only, never body.
  12. Hard numbers as design — specific figures over adjectives; data is the decoration in finance UI.
  13. State via environment — subtle bg tint to signal mode (live/paused, open/closed).
  14. Skeletons over spinners; instructional empty states (say what's missing + the next action).

8. Upgrading a ChatGPT/Grok-built project

  1. Classify the project type (§1) → load that skill + this one.
  2. Audit against §7. The usual offenders: system-ui font, no token layer, flat-gray everything, no focus states, default Tailwind blue everywhere, spinners, cramped or inconsistent spacing.
  3. Apply in order: fonts (§4) → token layer (§3) → spacing/grid → components to shadcn → motion → a11y. Fonts + tokens alone usually deliver ~70% of the perceived lift.
  4. Verify live on Vercel + run the §5 a11y checklist before calling it done.

Sources

Stripe/Vercel/Linear/Robinhood/Apple/Mercury/Arc teardowns; UX Planet "50 Design Styles"; WCAG 2.2 (w3.org); shadcn/Base UI vs Radix (2026); Motion.dev; Tailwind v4 docs; Typewolf/Google Fonts; utopia.fyi fluid type. Full per-track notes + URLs synthesized into this skill and the per-type skills. Treat ◐ as strong defaults; verify palettes with a contrast + CVD checker before shipping.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.