agentleFS
ryanschueryanschue/automated-product-engineer/skills/claude-design-to-app/SKILL.md

claude-design-to-app

Skill0 starsChanged 28 days agoMITView on GitHub

Prepare production-grade requests for Claude Design and turn Claude Design export folders into application UI. Use when a user asks what to send Claude Design, supplies a Claude Design handoff/export, or asks to implement or reconcile that export in an existing codebase. The skill covers the export contract, design-system mapping, real data and persistence, feature flags, analytics, responsive behavior, motion, accessibility, pixel comparison, and focused UI/E2E tests. Do not use for ordinary UI implementation without a Claude Design input, or for direct Figma implementation when only a Figma source is supplied.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 11 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Treat a Claude Design export as a design and behavior contract, not as production code to paste into the repository.

    line 8
  2. Distinctive:

    If the user wants a request to send to Claude Design, read claude-design-export-prompt.md, replace its bracketed inputs, and return the …

    line 19
  3. Distinctive:

    If the user supplied an export folder, read integration-checklist.md and follow it from discovery through verification.

    line 20
  4. Distinctive:

    If the user wants both, prepare the request first.

    line 21
  5. Distinctive:

    If the export points to a Figma source or the user also requests direct Figma implementation, invoke the applicable Figma design-to-code …

    line 22
  6. Distinctive:

    Use reasonable repository evidence for missing implementation context.

    line 35
  7. Distinctive:

    Start by inventorying the complete export.

    line 41
  8. Distinctive:

    Use real application data.

    line 50
  9. Distinctive:

    Keep dialog geometry stable across steps unless the design explicitly requires a size change.

    line 56
  10. Distinctive:

    Add meaningful focused coverage for:

    line 63
  11. Distinctive:

    Run broad repository gates only after design and behavior are stable.

    line 73

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Claude Design to App

Treat a Claude Design export as a design and behavior contract, not as production code to paste into the repository.

Adaptive intake

First infer product and repository context from the conversation, existing application, screenshots, and linked source. Compare what is known with the export contract below. If a missing product decision would change behavior, permissions, persistence, responsive rules, copy, analytics, or acceptance criteria, ask one to three focused questions instead of silently inventing it. Continue repository discovery while waiting when possible.

If the user needs Claude Design to supply missing design artifacts, generate one tailored copy-ready request using the prompt reference. Fill every field supported by current evidence, ask only for the missing decisions and deliverables, and explain what the returned package must prove. Do not present a menu of generic prompts. If the supplied context is sufficient, produce the request or begin integration without an intake ceremony.

Choose the workflow

  • If the user wants a request to send to Claude Design, read claude-design-export-prompt.md, replace its bracketed inputs, and return the complete prompt without weakening its deliverables.
  • If the user supplied an export folder, read integration-checklist.md and follow it from discovery through verification.
  • If the user wants both, prepare the request first. Integrate only after the resulting export is available.
  • If the export points to a Figma source or the user also requests direct Figma implementation, invoke the applicable Figma design-to-code skill before using Figma tools.

Requesting a Claude Design export

Collect or infer these inputs from the conversation and repository:

  1. Feature name and product context.
  2. Primary user and job to be done.
  3. Business rules, final copy, data definitions, persistence, permissions, analytics, and feature-flag rules.
  4. Known app design-system primitives, tokens, routes, and framework.
  5. Reference screenshots, files, existing flows, and required responsive sizes.
  6. Known product decisions and unresolved questions.

Use reasonable repository evidence for missing implementation context. Preserve unresolved product-policy choices as explicit decisions instead of silently inventing them.

Ask Claude Design for a self-contained handoff package with all meaningful states, deterministic state controls, typed reference components, separated copy and config, semantic tokens, local assets, machine-readable frame metadata, exact desktop and mobile reference images, accessibility and motion specifications, a data contract, analytics and persistence rules, acceptance criteria, and open decisions. The exact copyable request is in the prompt reference.

Implementing an export

Start by inventorying the complete export. Read the handoff, exact copy, configuration, tokens, source design, prototype, reference components, frame manifest, and canonical screenshots before editing the app.

Then inspect the target repository's instructions, design system, existing implementation, feature flags, data model, API patterns, analytics helpers, and tests. Write down a compact mapping from the export to existing primitives and semantic tokens. Resolve architecture and state ownership before broad visual work.

Use the export's reference component as an interaction and geometry model. Rebuild it with the app's components, tokens, routing, server authority, data access, and persistence conventions. Keep final copy and product thresholds/config out of presentation JSX when the app already has an appropriate configuration layer.

Preserve these contracts:

  • Gate every entry point and supporting endpoint that is part of a flagged feature. The flag-off path must have no visible or behavioral side effects.
  • Use real application data. Clearly label any fixture-only values and keep them out of production paths.
  • Persist completion or acknowledgment server-side when the experience must survive devices or sessions.
  • Distinguish first-time, replay, reminder, and existing-user flows. Do not let replay mutate completion state or redirect unexpectedly.
  • Separate intermediate persistence from final close/navigation. A refresh caused by an intermediate save must not make the current experience ineligible and unmount its final screen.
  • Define dismissal, Escape, outside-click, failure recovery, and navigation behavior deliberately for each flow.
  • Track initial open, step views, dwell time, actions, completion, dismissal, failure, and replay when analytics are requested.
  • Keep dialog geometry stable across steps unless the design explicitly requires a size change.
  • Implement reduced-motion behavior alongside the intended transitions.

Verify the result

Create deterministic fixtures or URL controls for each canonical state when the repository permits them. Capture the exact viewports listed in the export's frame manifest and compare them against the supplied references with overlays or image diffs. Inspect desktop and mobile manually; fix colors, type, radii, spacing, alignment, wrapping, overflow, and motion before accepting visual baselines.

Add meaningful focused coverage for:

  • eligibility and feature-flag isolation;
  • state transitions, guards, boundary values, and replay behavior;
  • persistence/API success and failure;
  • dismissal and keyboard behavior;
  • analytics payloads and timing lifecycle;
  • responsive geometry and canonical visual states;
  • final navigation and server-refreshed state.

Run broad repository gates only after design and behavior are stable. Report what changed, how it maps to the export, the visual evidence, tests run, and any product decisions that remain open.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.