What problem does it solve? Turning a visual reference — a screenshot, a live website, or a Figma file — into a documented, machine-readable design system normally requires hours of manual inspection and guesswork. This Skill automates that analysis, producing a structured design.md spec, W3C DTCG design tokens, and optional WCAG contrast reports with confidence markers on every inference. ## Core Features & Use Cases - Full design-system extraction: Analyzes images, URLs (HTML + CSS custom properties + Playwright screenshots), and Figma files via MCP to produce a 7-section design.md with tokens, components, layout, brand rules, and open questions. - Element mode: Copies a single element (navbar, card, 3D illustration) into a focused element.md — either a rebuild prompt for code agents like v0 and Claude Code, or a token-grounded image-generation prompt for gpt-image, Midjourney, or Stable Diffusion. - Standalone CLI scripts: Seven Python tools for CSS variable extraction, multi-viewport capture, dominant color extraction, WCAG contrast checking, design.md linting, token drift auditing against live sites, and export bundles for claude.ai/design. - Use Case: Point it at vercel.com to extract 808 CSS custom properties into DTCG tokens, then paste the resulting design.md into v0 to rebuild the page, or run verify_design.py months later to check whether the spec still matches the live site. ## Quick Start Ask the AI to extract the design system from a website URL, screenshot, or Figma link you provide, and it will generate a design.md with tokens and reconstruction notes.