DESIGNLANG — Design System Extraction
Skill by ara.so — Daily 2026 Skills collection.
designlang crawls any website with a headless browser (Playwright), extracts every computed style from the live DOM, and generates 8 output files: AI-optimized markdown, visual HTML preview, Tailwind config, React theme, shadcn/ui theme, Figma variables, W3C design tokens, and CSS custom properties.
Installation
Core Commands
Basic Extraction
Specialized Extraction
Authentication
Utility Commands
Output Files
Running npx designlang https://stripe.com produces:
Full CLI Reference
Common Patterns
Pattern 1: Extract and apply to a Tailwind project
Generated tailwind.config.js looks like:
Pattern 2: Feed the markdown to an LLM
Pattern 3: Compare competitors
Pattern 4: Apply to a shadcn/ui project
Generated shadcn-theme.css:
Pattern 5: W3C Design Tokens for tooling
Generated design-tokens.json:
Pattern 6: React theme object
Generated theme.js:
Pattern 7: Monitor a design system for changes
Pattern 8: Clone a site's design as a Next.js starter
What Gets Extracted
The markdown output covers 19 sections:
- Color Palette — all unique colors with usage context
- Typography — font families, weights, sizes, line heights, letter spacing
- Spacing — padding/margin/gap values used across the site
- Border Radii — all radius values with component context
- Box Shadows — elevation system
- CSS Custom Properties — all
--vardeclarations - Breakpoints — detected responsive breakpoints
- Transitions & Animations — duration, easing, properties animated
- Component Patterns — button, card, input, nav with full CSS snippets
- Layout System — grid columns, flex patterns, container widths
- Responsive Design — what changes across breakpoints (with
--responsive) - Interaction States — hover/focus/active deltas (with
--interactions) - Accessibility — WCAG 2.1 contrast scores for all fg/bg pairs
- Gradients — type, direction, stops, classification
- Z-Index Map — layer hierarchy, z-index wars detection
- SVG Icons — deduplicated, size/style classified, color palette
- Font Files — source (Google/self-hosted/CDN/system),
@font-faceCSS - Image Style Patterns — aspect ratios, filters, shape treatments
- Quick Start — ready-to-use snippet to recreate the design
Design Scoring
Scores 7 categories on a 0–100 scale with A-F grade:
- Color Discipline — palette size, harmony, usage consistency
- Typography — scale rationality, weight/size combinations
- Spacing System — whether spacing follows a scale
- Shadows — elevation system coherence
- Border Radii — consistency across components
- Accessibility — WCAG 2.1 contrast pass rate
- Tokenization — CSS custom property usage
Troubleshooting
Site loads blank or styles are missing
Authentication-protected pages
Output files are too large
Too many colors extracted (utility CSS sites)
Playwright/browser not found
apply command doesn't detect my framework
Integration with AI Agents
The -design-language.md file is specifically structured for LLM consumption. Feed it directly to Claude, GPT-4, or Cursor to:
- Recreate UI components matching the extracted design
- Generate new components consistent with the design system
- Audit your own components against a reference design
- Document your own site's design system automatically

