awesome-claude-design (rohitg00) — DESIGN.md Prompts by Aesthetic Family

Source: Rohitg00 Awesome Claude Design 2026 04 19

Repo: https://github.com/rohitg00/awesome-claude-design Stars: 95 (at time of ingest 2026-04-19) License: MIT Created: 2026-04-19 (two days after Claude Design launched)

Awesome-list curated around Anthropic’s Claude Design (Anthropic Labs) product. Collects DESIGN.md prompts organized by aesthetic family (not industry), remix recipes, five prompt packs, video teardowns, X/Twitter launch reactions, and honest community commentary. Complements VoltAgent’s awesome-design-md: VoltAgent covers the general DESIGN.md format across all AI agents; this repo is Claude-Design-specific with workflow, prompt, and community-signal layers on top.

Key Takeaways

• Aesthetic-family taxonomy, not industry. Nine families: Editorial Minimalism (Linear/Stripe/Vercel), Terminal-Core (Ollama/Warp/Raycast), Warm Editorial (Claude/Anthropic/Notion), Data-Dense Pro (ClickHouse/PostHog/Grafana), Cinematic Dark (RunwayML/ElevenLabs/Midjourney), Playful Color (Figma/Clay/Duolingo), Glass/Soft-Futurism (Apple/Arc/Granola), Neon Brutalist (The Verge/PlayStation/Pitchfork), Cult/Indie Picks. Aesthetic-first organization composes better with “pick a vibe” briefs than industry buckets.
• Remix recipes — blend two brands. Token combinations like Linear × Claude or Warp × Sentry. Concrete mechanism for generating novel aesthetics rather than cloning one brand. Pairs with the “remix-two-brands” prompt pack.
• Five production prompt packs. brand-to-design-md, audit-live-site, 3-designer-debate, remix-two-brands, family-picker. Effectively a starter toolkit for Claude Design users — not just the design-system file.
• Eight video teardowns aggregated. Malewicz, 02ui, WorldofAI, Ray Fernando, Greg Isenberg, et al. Useful for calibrating expectations before burning Claude Design usage.
• Skills + MCP integration notes. Lists design-shotgun, design-html, design-review skills; SuperDesign MCP integration for Claude Code.
• Community signal captured. Launch-day X reactions, stock-market impact (Figma closed −4.26% on launch day), professional-designer pushback — material for positioning calls when recommending Claude Design to stakeholders.
• Consensus takeaways flagged. Wins: design-system coherence, web capture, collaboration. Losses: “burns through usage quickly,” hallucination on long tasks, Opus 4.7 required for best output.
• Fresh and moving fast. Created same day as this ingest (2026-04-19); 95 stars within hours; single maintainer (rohitg00); MIT.

How it composes with the design-skills stack

• vs VoltAgent awesome-design-md: VoltAgent’s collection targets any AI agent that reads a DESIGN.md (Claude Code, Cursor, Google Stitch) and is organized site-by-site. This repo is scoped to Claude Design and organized aesthetic-family-first — a workflow layer on top of the DESIGN.md format.
• vs Design Inspiration Galleries: Galleries are the references humans browse before prompting. This repo is prompts + remix recipes — the layer between reference-gathering and generation.
• vs Anti-AI Slop Guide: The repo includes an “Anti-Slop Kit” — complementary vocabulary for a well-documented problem. Worth cross-reading for additional banned patterns.
• vs Taste Skill Deep Dive: Aesthetic families map loosely to Taste’s design dials and vibe archetypes. Useful cross-reference when picking a family.
• vs Soft Skill Creative Direction: Soft Skill’s vibe archetypes and Absolute Zero directive are methodology; aesthetic families are concrete target specs. Same problem, opposite ends of the abstraction axis.

Implementation

• Tool/Service: rohitg00/awesome-claude-design (GitHub awesome-list)
• Setup:
  1. Visit https://github.com/rohitg00/awesome-claude-design
  2. Pick an aesthetic family matching your brief
  3. Grab the relevant DESIGN.md prompt and paste into Claude Design (or use via Claude Code with a DESIGN.md checked into the repo)
  4. For remixes: pick two brand tokens, use the remix-two-brands prompt pack
  5. For existing sites: use the audit-live-site prompt pack to extract tokens into DESIGN.md format
• Cost: Free (MIT). Running the prompts in Claude Design counts against your plan’s Claude Design usage — per community signal, “burns through usage quickly” on long tasks.
• Integration notes:
  • Opus 4.7 required for best output (per community consensus captured in the repo).
  • Pairs with SuperDesign MCP for IDE-native workflows in Claude Code.
  • Can be cloned as a submodule into any repo that generates UI via Claude Code — the DESIGN.md files and prompts become reusable assets.

Related

• Claude Design (Anthropic Labs) — the Anthropic product this repo is entirely organized around
• awesome-design-md (VoltAgent) — the general DESIGN.md collection; this repo is the Claude-Design-specific companion
• Design Inspiration Galleries — reference-gathering layer; aesthetic families let you translate gallery picks into DESIGN.md prompts
• Tutorial: Claude Design for Prototypes and UX — workflow context for where these prompts slot in
• Tutorial: Claude Design for Presentations and Slide Decks — slide-deck side of Claude Design; family prompts apply
• Anti-AI Slop Guide — the Anti-Slop Kit in this repo composes with the guide’s banned-pattern list
• Frontend Design Deep Dive — Claude’s master design skill; aesthetic families are target specs to reason from
• Taste Skill Deep Dive — design dials and vibe archetypes; aesthetic families map to these
• Soft Skill Creative Direction — Absolute Zero directive; composes with the family-picker prompt
• Claude Prompting Best Practices — the “specify a concrete alternative” guidance; aesthetic families are concrete alternatives

Open Questions

• Single maintainer, fresh repo. Active as of 2026-04-19 but one contributor. Durability risk — worth monitoring over the next 60 days to see whether it becomes a canonical community hub or stalls.
• Overlap with Anthropic’s own prompt library. 40+ official prompts already exist at Anthropic’s prompt library. Which prompts here extend beyond what’s official, and which duplicate? Not yet audited.
• Video teardown quality. Eight teardowns listed — which are worth the time investment? No community rating in the repo.
• Remix-recipe fidelity. Blending two brand tokens sounds powerful in principle. In practice, how coherent are the outputs? Empirical testing not yet done.
• Anti-Slop Kit vs our own Anti-AI Slop Guide. Delta between the two deserves a reconciliation pass.
• Claude Design “burns usage quickly.” Community takeaway captured but unquantified — how many generations per free-tier session before hitting limits? Worth measuring directly.

Try It

1. Pick a family matching an in-progress brief. Open the repo, scan the nine aesthetic families, pick one. Screenshot the family’s reference sites. Use the family-picker prompt to confirm the match before committing.
2. Try a remix recipe on a real brief. Pick two brands whose tokens should combine well (e.g., Linear × Anthropic for a dev-focused editorial vibe). Run remix-two-brands. Compare the output to picking one brand alone.
3. Audit your own site into a DESIGN.md. Run audit-live-site on your current brand. Compare the extracted tokens to what you thought your brand system was. Useful diagnostic even if you never regenerate.
4. Watch the 3-designer-debate prompt pack. Feeding a single brief through three designer personas produces more differentiated output than a single-designer prompt. Effectively a local version of the Advisor Strategy for design work.
5. Skim two video teardowns. Before committing Claude Design to a real project, spend 20 minutes on the Malewicz and Greg Isenberg teardowns to calibrate expectations against reality.
6. Track the repo. 95 stars same day as this ingest; worth watching over the next few weeks. Star + subscribe to releases.
7. Cross-reference the Anti-Slop Kit. Read both this repo’s Anti-Slop Kit and our own Anti-AI Slop Guide — note any patterns one covers that the other misses, file back if material.