agentleFS
ForgeyClapForgeyClap/claude-forge/.claude/skills/forge-figma/SKILL.md

forge-figma

Skill2 starsChanged 35 days agoMITView on GitHub

Forge: design-to-code: figma, design to code, mockup, design tokens, pixel-perfect, responsive, breakpoints, style guide, dev mode, handoff, accessibility.

1 other project uses a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Forge playbook — Design → code (Figma / mockups)

ui-boss leads the translation under build-boss; a stack reviewer (react-reviewer or vue-reviewer) reviews the built components; a11y-architect guards accessibility. Design→code is website/frontend work, so the built result also runs under forge-website conventions (real content, responsive, a11y, no placeholders). The honesty rule here: state what you translated from — a live Figma read (MCP) or provided exports — and never invent measurements you couldn't see.

Hard rules (non-negotiable)

  • Faithful translation, verified visually. Rendered output is compared side-by-side against the source frame at the design's breakpoints (spacing, sizing, hierarchy, states) — "close enough" from memory is not translation. Screenshot evidence, not assertion.
  • Design tokens, not scattered magic values. Colours, typography, spacing, radii, shadows are extracted into tokens/variables (CSS custom properties, a tokens file, or a theme object) and reused — the design's real scale, not a fresh hardcoded hex/px per component.
  • Responsive fidelity. The design's breakpoints are honoured (Figma usually ships desktop + mobile frames); layout adapts. No fixed-pixel-only layout that breaks off-frame.
  • Accessibility is not optional. Semantic HTML, WCAG AA contrast, visible focus states, keyboard navigation, alt text, ARIA where needed. A design can look right and still fail contrast — flag the failure back, don't silently ship it. a11y wins over exact pixel-match.
  • Component structure mirrors the design system. A Figma variant set (e.g. a Button with variants) becomes one prop-driven component, not copy-pasted one-offs. Real content from the design, no Lorem where real copy exists.

Team (conditional by stack/level)

Lead: ui-boss (design-led) or build-boss. Build review: react-reviewer or vue-reviewer (by stack). Accessibility: a11y-architect. Figma MCP wiring (opt-in): mcp-developer. Design critique of the translated result: the design-is skill. If it's a marketing/landing surface, seo-boss for meta/perf.

Skills / commands / MCP

design-is (audit the translated UI against real design principles — see ~/.claude/rules/ecc/web/design-quality.md), browser (screenshot loop: render → compare to the source frame at each breakpoint), forge-website (the built UI must be responsive + a11y + real-content), /react-build + /react-review (or the Vue equivalents). Figma MCP is opt-in: the Figma Dev-Mode MCP server / Figma API lets an agent read the file and pull tokens, measurements, and exports directly — it requires a Figma access token + MCP wiring (mcp-developer sets it up), and this session's n8n/other MCP auth state does not include it by default. Native fallback (no MCP): work from owner-provided exports — PNG/SVG frames + a measurement/redlines spec (or a shared style guide). Be explicit: without the MCP or provided exports, Claude cannot see the Figma file and any translation is approximate — request the assets rather than guessing.

Fan-out & flow

L2 a single component or screen; L3 a full design-system / multi-screen translation. Serial: ingest the design (MCP read or exported frames + specs) → extract tokens → build components → wire responsive breakpoints → a11y pass → visual diff vs source. Parallel: independent components/screens translate in parallel once the token system is fixed (tokens are the shared contract).

Domain gates

  • Rendered output matches the source frame at the design's breakpoints — side-by-side screenshot evidence.
  • Design tokens extracted to variables/theme and reused; no per-component hardcoded palette/spacing drift.
  • Responsive at the design's declared breakpoints; no fixed-pixel-only layout.
  • a11y: contrast meets AA, semantic HTML, visible focus, keyboard-navigable — any contrast failure in the source is flagged, not shipped silently.
  • Components mirror the design's variant/component system (reusable, prop-driven), not copy-paste duplicates.
  • The translation source is stated — Figma MCP read vs provided exports — and no measurements are invented.

Ship-readiness (unique)

Side-by-side visual match at the design's breakpoints (screenshots attached); a tokens file/theme present and used; responsive verified; a11y (contrast AA / focus / keyboard) checked with any source-side a11y defects flagged back; component structure mirrors the design system; the translation source (MCP vs exports) stated honestly — and if neither was available, the output is labelled approximate. Defers to forge-website ship-readiness for the built page (real content, no test copy, secrets in env). Advisory checklist; optionally run codex-reviewer on component code — not a blocker.

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.