Skip to content
cccskills
FreePublished on GitHub

design-system-generator

Generates production-ready design tokens from natural language descriptions. Outputs Tailwind configs, CSS custom properties, and DTCG W3C tokens from 24 trends + 31 AI-ready styles with 2800+ component examples.

The English summary is being prepared. Showing the original description.

See how to install

Files included (2)

  • SKILL.md8.4 KB
  • scripts/match-trend.ts6.4 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Design System Generator

Transform natural language descriptions into production-ready design tokens. Powered by 24 design trends, 31 AI-ready styles, and 2,800+ component examples.

DECISION POINTS

1. Confidence Threshold Rules

If trend match confidence >= 80%:
  → Generate immediately with primary trend

If confidence 60-79%:
  → Show top 2 matches, ask user to choose
  → Example: "Swiss-modern (65%) vs. Hyperminimalism (62%) - which feels right?"

If confidence < 60%:
  → Use vibe-matcher skill first for emotional clarity
  → Re-run trend matching with vibe context

2. Multi-Trend Disambiguation

If 2+ trends score within 10% of each other:
  ├─ Business/professional context?
  │   → Prefer: swiss-modern, hyperminimalism, dark-mode
  ├─ Creative/agency context?
  │   → Prefer: neobrutalism, maximalism, experimental-navigation
  ├─ Tech/developer context?
  │   → Prefer: terminal-aesthetic, glassmorphism, web3-crypto
  └─ Consumer app context?
      → Prefer: claymorphism, botanical-organic, gamified-design

3. Output Format Selection

If user mentions specific framework:
  ├─ "Next.js", "React", "Vite" → generate-tailwind.ts
  ├─ "Vue", "Angular", "vanilla" → generate-css-vars.ts
  └─ "Figma", "Style Dictionary" → generate-tokens.ts (DTCG)

If no framework mentioned:
  → Ask: "What's your tech stack?" or generate all three formats

4. Accessibility Requirements

If trend has accessibility warnings (glassmorphism, claymorphism):
  → Generate standard tokens first
  → Run contrast checker
  → If violations found:
      ├─ Adjust color values to meet WCAG AA
      ├─ Document overrides in comments
      └─ Warn user about visual changes

5. Scope Determination

If user mentions theming:
  ├─ "dark mode" → Generate with --scope=.dark
  ├─ "multiple brands" → Generate with --scope=[data-theme="name"]
  └─ "component library" → Use :root scope (default)

If existing design system mentioned:
  → Generate with theme.extend structure only

FAILURE MODES

1. Wrong Color Direction (Vibrant when user wanted muted)

Detection: User says "too bright", "too colorful", "more subtle" Symptoms: High saturation values, neon/electric colors in conservative context Fix: Re-run with lower-energy trend (swiss-modern instead of neobrutalism)

# Wrong
npx ts-node generate-tailwind.ts neobrutalism  # Generates #FF5252, #00E676

# Fixed  
npx ts-node generate-tailwind.ts swiss-modern  # Generates #374151, #6B7280

2. Compilation Chaos (Generated tokens break build)

Detection: Tailwind purge warnings, CSS custom property undefined errors Symptoms: Missing quotes in font names, invalid CSS values, circular references Fix: Validate output before deployment

# Check Tailwind compilation
npx tailwindcss -i ./src/input.css -o ./dist/output.css --watch

# Check CSS syntax
npx stylelint tokens.css

3. Contrast Violation (Accessibility failure in production)

Detection: Lighthouse audit < 90, WAVE tool errors Symptoms: Text invisible on backgrounds, pale borders, low-contrast buttons Fix: Run contrast audit before finalizing

# Audit generated tokens
npx ts-node scripts/check-contrast.ts swiss-modern
# → Warning: text-gray-400 on bg-white = 2.1:1 (needs 4.5:1)

4. Semantic Mismatch (Token names don't match visual result)

Detection: Developer confusion, "primary" color isn't prominent Symptoms: brutal-red is actually orange, glass-white is gray Fix: Check trend mapping accuracy

# Verify semantic alignment
npx ts-node scripts/preview-tokens.ts neobrutalism
# Opens browser with all colors labeled

5. Scope Pollution (CSS variables leak across themes)

Detection: Theme switching doesn't work, colors persist incorrectly Symptoms: Dark mode shows light colors, component themes interfere Fix: Use explicit scoping

/* Wrong - leaks to global */
:root { --color-primary: #FF0000; }

/* Fixed - scoped properly */
[data-theme="brand-a"] { --color-primary: #FF0000; }
[data-theme="brand-b"] { --color-primary: #00FF00; }

WORKED EXAMPLES

Scenario: SaaS Dashboard for Developer Tools

Input: "Clean dashboard for developers, needs to feel professional but not boring"

Step 1: Trend Matching

npx ts-node scripts/match-trend.ts "clean dashboard for developers professional"

Result: swiss-modern (78%), terminal-aesthetic (45%)

Decision: 78% confidence is good, but "not boring" suggests terminal-aesthetic could add personality. Check both.

Step 2: Preview Both Options

  • Swiss-modern: #374151 grays, Helvetica fonts, subtle shadows
  • Terminal-aesthetic: #00E676 greens, monospace, sharp edges

Expert insight: For developer tools, terminal-aesthetic creates better brand differentiation while maintaining professionalism.

Step 3: Generate Tailwind Config

npx ts-node generate-tailwind.ts terminal-aesthetic > tailwind.tokens.ts

Step 4: Quality Check

// Generated output includes:
colors: {
  'terminal-green': '#00E676',
  'terminal-dark': '#0D1117', 
  'terminal-gray': '#21262D'
}

Step 5: Accessibility Verification

  • terminal-green on terminal-dark = 7.2:1 ✅ (exceeds AAA)
  • All text combinations pass WCAG AA

Result: Professional developer aesthetic with personality, fully accessible.

QUALITY GATES

Pre-deployment checklist for generated tokens:

  • WCAG Contrast: All text/background combinations ≥ 4.5:1 (AA) or ≥ 3:1 for large text
  • Tailwind Compilation: npx tailwindcss builds without warnings or errors
  • CSS Validation: Custom properties have valid syntax (no missing semicolons, quotes)
  • Browser Compatibility: Tokens render correctly in Chrome, Firefox, Safari latest versions
  • Semantic Accuracy: Color names match visual appearance (red looks red, not orange)
  • Component Coverage: Generated tokens support primary UI components (buttons, inputs, cards)
  • Theme Isolation: Multiple themes don't interfere (test theme switching if applicable)
  • Font Loading: Generated font families are available or have fallbacks
  • Performance: CSS file size < 50KB, no unused custom properties
  • Documentation: Token purpose and usage examples are clear to implementers

NOT-FOR BOUNDARIES

❌ DO NOT use for:

  • Color exploration/mood boarding → Use vibe-matcher skill instead
  • Existing design system refinement → Use design-system-documenter for optimization
  • Component code generation → Use component-template-generator with these tokens
  • Brand identity creation → Use color-theory-palette-harmony-expert for strategic color work
  • Single token adjustments → Just edit the values directly
  • Complex animation systems → Use web-design-expert for motion design
  • Marketing site aesthetics → This generates systematic tokens, not one-off visual styles

✅ Perfect for:

  • New project foundation setup
  • Converting design descriptions to code
  • Standardizing ad-hoc color usage
  • Cross-format token consistency (Tailwind + CSS vars + DTCG)

Delegation Rules:

  • If user needs emotional/conceptual clarity → vibe-matcher first, then return here
  • If user has existing system → design-system-documenter for analysis
  • If user needs components, not tokens → component-template-generator with generated tokens
  • If user needs color strategy → color-theory-palette-harmony-expert for theory, then return here

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Expert in 2000s-era music visualization (Milkdrop, AVS, Geiss) and modern WebGL implementations. Specializes in Butterchurn integration, Web Audio API AnalyserNode FFT data, GLSL shaders for audio-reactive visuals, and psychedelic generative art. Activate on "Milkdrop", "music visualization", "WebGL visualizer", "Butterchurn", "audio reactive", "FFT visualization", "spectrum analyzer". NOT for simple bar charts/waveforms (use basic canvas), video editing, or non-audio visuals.

The English summary is being prepared. Showing the original description.

curiositech/port-daddy3Updated Oct 19, 2026

Expert legal research agent for finding and scraping expungement data state by state. Knows authoritative sources, URL patterns, Firecrawl configuration, and 2026 legal landscape.

The English summary is being prepared. Showing the original description.

curiositech/port-daddy3Updated Oct 19, 2026

Expert in 3D computer vision labeling tools, workflows, and AI-assisted annotation for LiDAR, point clouds, and sensor fusion. Covers SAM4D/Point-SAM, human-in-the-loop architectures, and vertical-specific training strategies. Activate on '3D labeling', 'point cloud annotation', 'LiDAR labeling', 'SAM 3D', 'SAM4D', 'sensor fusion annotation', '3D bounding box', 'semantic segmentation point cloud'. NOT for 2D image labeling (use clip-aware-embeddings), general ML training (use ml-engineer), video annotation without 3D (use computer-vision-pipeline), or VLM prompt engineering (use prompt-engineer).

The English summary is being prepared. Showing the original description.

curiositech/port-daddy3Updated Oct 19, 2026

Apply crisis decision-making research to agent routing, uncertainty triage, and coordination failure analysis in time-pressured systems. Use when diagnosing handoff failures, analytical paralysis, or expert judgment under incomplete information. NOT for routine coding, simple CRUD design, or static single-agent tasks with complete information.

The English summary is being prepared. Showing the original description.

curiositech/port-daddy3Updated Oct 19, 2026

Use for insight, reframing, contradiction, impasse, and anomaly-driven problem solving when execution effort no longer helps. NOT for routine optimization, error correction, or well-specified tasks with known solution paths.

The English summary is being prepared. Showing the original description.

curiositech/port-daddy3Updated Oct 19, 2026

Apply cognitive task analysis to expert work that depends on perceptual cues, branching judgment, and recurring monitoring loops. Use when decomposing expert capability into agent structure, simulation design, or validation interviews. NOT for ordinary step-by-step SOP capture or simple pipelines with no tacit cue layer.

The English summary is being prepared. Showing the original description.

curiositech/port-daddy3Updated Oct 19, 2026

See all skills by curiositech

Report a problem with this skill