What it does This repo collects DESIGN.md files—plain-text design system documents that describe how popular sites look and feel. Drop one into your project, point an AI coding agent at it, and the agent generates UI that matches the described aesthetic: colors, typography, spacing, and interaction patterns.

The format comes from Google Stitch. No JSON schemas, no exports, no plugins. Just markdown, because LLMs already read markdown natively.

The interesting bit The value isn’t the format—it’s the analysis. Each file extracts the actual design logic from real sites, down to specific tokens like “112px display type” (Sanity) or “acid-mint and ultraviolet accents” (The Verge). It’s design-system archaeology performed so your agent doesn’t have to squint at a screenshot and guess.

Key highlights

• 72 DESIGN.md files covering brands from Apple and Spotify to Binance and SpaceX
• Categories span AI platforms, dev tools, fintech, e-commerce, automotive, and media
• Files include analyzed patterns, tokens, and rules rather than surface-level descriptions
• Request service available for private or custom site analysis
• Companion ecosystem section promotes related AI build tools (notably VoltAgent, the repo’s parent project)

Caveats

• The actual DESIGN.md files live on an external site (getdesign.md); this repo is essentially an index and landing page
• README is upfront about being “ranked #150 globally on GitHub” and actively seeks sponsors—useful context, not a neutral curator
• No visible verification of how faithfully the generated files reproduce the original designs

Verdict Grab this if you’re vibe-coding landing pages and want your AI to stop generating generic Tailwind defaults. Skip it if you need rigorous, battle-tested design tokens for production systems—this is shortcut fuel, not infrastructure.