DESIGN.md releases 500+ remixable specs and a 43-minute motion workflow
MengTo published a 43-minute workflow and 500+ remixable DESIGN.md files for turning specs into landing pages, mobile screens, and motion concepts. Use the library as reusable design memory across Stitch, Claude Design, and code handoffs.

TL;DR
- MengTo's resource thread points to 500+ remixable DESIGN.md files on getdesign.md plus a 43-minute workflow for turning one spec into landing pages, mobile screens, and motion concepts.
- In MengTo's takeaways thread, DESIGN.md is framed as reusable project memory: one markdown file for colors, type, spacing, and component rules that can move between Stitch, coding agents, repos, and builders.
- 7_eito_7's feature rundown says the draft spec already exposes a Tokens section, a Components section in progress, and a CLI validator for catching spec mistakes.
- According to the HN discussion highlights, people are already using Claude Design for layout exploration, then handing the result to Claude Code or backend tools for implementation.
You can browse the DESIGN.md repo, check Anthropic's Claude Design announcement, and steal directly from getdesign.md. The useful bit in MengTo's workflow is not the demo polish, it's the pipeline: start with a spec, generate a first direction, remix it, expand it into more screens, then pass the system into a builder or code handoff MengTo's resource thread MengTo's takeaways thread.
500+ remixable specs
MengTo's post is basically a starter pack for designers who do not want to prompt from scratch every time.
The bundle ties together five surfaces:
- DESIGN.md GitHub for the draft spec.
- Stitch for generation and remixing.
- getdesign.md for the 500-plus reusable files.
- Aura for building and publishing.
- Claude Design as another design generation surface.
In MengTo's takeaways thread, the workflow order is unusually explicit: start with DESIGN.md, generate the first design, remix and expand it, create section variants, then move into a builder to assemble the full site. That is a cleaner handoff story than the usual screenshot-plus-prompt soup.
DESIGN.md as shared memory
The core pitch is simple: move design rules out of scattered tools and into a markdown file agents can actually parse.
According to MengTo's takeaways thread, that middle layer works because markdown is structured enough for agents but still readable for designers and developers. 7_eito_7's thread post makes the same point more bluntly, listing the old state as Figma, Notion, and docs split apart, then collapsing color, typography, components, and rules into one file.
The draft spec pieces called out in the evidence are concrete:
- Tokens: centralized design variables, per 7_eito_7's feature rundown.
- Components: a section for UI part definitions, still in progress, per 7_eito_7's feature rundown.
- CLI Validator: automated checks for spec mistakes, per 7_eito_7's feature rundown.
- Accessibility and linting: MengTo says a DESIGN.md file can be checked for missing essentials, broken references, section order, and contrast issues in his takeaways thread.
That last part is the sleeper feature. A portable style guide is useful, but a style guide that can be linted before the UI ships is much more interesting.
Claude Design handoff
The community angle is less about pure mockups and more about collapsing the distance between mockup and working product.
Discussion around Claude Design
1.2k upvotes · 759 comments
The Hacker News thread collects three concrete patterns:
- One commenter used Claude Design to explore layouts, tweak details with comments, and export the design system into code, according to the florakel comment.
- Another prototyped a CRUD app in Claude Design, then handed it to Claude Code inside an existing business stack, according to the DecoPerson comment.
- A third shared a two-line prompt that pulls an InstantDB AGENTS.md URL to spin up a backend-backed app flow, according to the stopachka comment.
Claude Design
1.2k upvotes · 759 comments
That makes DESIGN.md look less like a file format story and more like a handoff protocol. MengTo's own list in his takeaways thread lands in the same place: a good hero should branch into pricing, testimonials, mobile layouts, motion, branding boards, and then implementation, without losing the original system.