amnafarzy/ravie/figma-lovable-handoff
Claude Code and agent skill by amnafarzy/ravie
- community
Checking that the SKILL.md file is still in the repository.
Use this whenever converting a Figma frame or Lovable prototype into an implementation plan — component architecture, token map, states, QA criteria — even if the user just pastes a design link and says "build this". Requires an accessible design reference.
Install
npx vskill@latest install amnafarzy/ravie/figma-lovable-handoffSKILL.md
The same SKILL.md was published earlier as amnafarzy/ravie/figma-lovable-handoff.
figma-lovable-handoff
Convert designs from Figma or prototypes from Lovable into production-ready implementation. Critical for designer-developer workflow.
Purpose
Take a design source (Figma file, screenshot, Lovable prototype URL) and produce: a token map, component spec, state map, accessibility requirements, Linear issues, and an implementation plan that fits your actual stack — without copying garbage code.
Stack integration
Reads from:
- Figma (file, screenshots, design tokens, comments)
- Lovable (prototype URL, generated code if accessible)
- Notion (PRD with user goal context)
- GitHub (existing component library, design system tokens, conventions in
CLAUDE.md) - Linear (acceptance criteria for the implementation issue)
Writes to:
- Implementation spec (in chat, then to Notion if durable)
- Linear issues (if approved, broken into per-component tasks)
- GitHub plan (in repo if structured spec is needed in repo)
- QA checklist for
vercel-preview-qato use later
Inputs required
- Design source URL or attached screenshot
- Target repo and component location
- Product goal (from Notion PRD or stated by user)
- Breakpoints to support (default: mobile 375, tablet 768, desktop 1280)
- States to handle (loading, empty, error, success, disabled, hover, focus)
- Existing tokens to reuse (from repo's design system)
- Lovable code disposition: keep / rewrite / reference / discard
Process
0. Read the UI rule reference first
Before UI work or review, read rules/ui.md and apply the project’s responsive, accessibility, component-state, and design-token expectations.
1. Identify product intent (not just visuals)
The design shows what; you need to extract why:
- What is the user trying to accomplish on this screen?
- What's the most important action?
- What can be removed without breaking the intent?
If the PRD doesn't answer this, route to requirements-griller first.
2. Extract design tokens
Pull from the Figma file or infer from screenshot:
- Colors: hex values for backgrounds, text, borders, accents
- Typography: font family, size, weight, line height per text style
- Spacing: padding, margin, gap values used
- Radius: border-radius values
- Shadows: any used (note: stack discourages shadows; flag if design relies on them)
- Motion: any specified animation duration/easing
Map each token against your existing design system in the repo. Output:
If the design introduces new tokens, flag them for explicit approval before adding. Don't silently expand the token system.
3. Map screens to components
Decompose the design into component hierarchy:
For each component:
- Existing? Reuse
- Variant of existing? Add variant
- New? Build new with existing primitives
- Lovable-generated? Mark for rewrite
4. Identify all states
For every interactive component:
- Default
- Hover
- Focus (keyboard)
- Active / pressed
- Disabled
- Loading
- Empty (if applicable)
- Error
- Success
If the design only shows the default state, document the assumption that defaults apply (matching design system) — flag for designer review.
5. Define responsive behavior
Default breakpoints (override from CLAUDE.md if project specifies different):
- Mobile: 375px
- Tablet: 768px
- Desktop: 1280px
For each breakpoint, document:
- Layout direction (stack / row / grid)
- Spacing changes
- Hidden / shown elements
- Font size adjustments
6. Define accessibility requirements
Apply the project's a11y baseline (from CLAUDE.md or the ui-quality skill):
- Semantic HTML elements required
- ARIA labels needed
- Keyboard navigation expected
- Focus order
- Touch target minimum (44×44 default)
- Reduced-motion alternatives
- Contrast ratios
7. Identify Supabase / data dependencies
What data does this UI need?
- Existing tables / queries
- New fields needed
- Auth state requirements
- RLS policy implications
If new data shape needed, flag a sub-task for supabase-guardian review.
8. Triage Lovable code (if applicable)
For each generated file/component:
- Keep: matches conventions, no regressions
- Rewrite: same shape, but use your tokens / patterns / data layer
- Reference: useful as inspiration, don't copy directly
- Delete: unhelpful, add nothing
Default disposition: rewrite unless code is trivial. Lovable's output is good for prototyping, almost never production-ready as-is.
9. Produce implementation plan
10. Create QA checklist for the preview
A prepared checklist for vercel-preview-qa to run later:
11. Handoff
Stop before implementation. Output the spec and wait for approval to:
- Create Linear issues from the plan
- Add new tokens (if any) — separate approval required
- Begin implementation (handoff to
issue-to-pr)
Output format
Approval gates
Tier 1-2 (default) — extracting spec, drafting plan: no approval needed.
Tier 3 — needs approval:
- Adding new design tokens to the system
- Creating Linear issues from the plan
- Committing the spec to GitHub
Never auto:
- Copying Lovable-generated code into production paths
- Changing existing tokens
- Removing existing component variants
- Adding heavy new dependencies (animation libs, chart libs, etc.)
Hard rules
- Never copy Lovable code verbatim into production — always rewrite or explicitly mark as kept
- Never silently expand the token system — propose, get approval, then add
- Never implement only the happy path — every state must be defined
- Never skip the responsive pass — mobile is not a squeezed desktop
- Never invent exact pixel values — note assumption when unclear
- Never skip the accessibility requirements section — even if "we'll add it later"
Connects to
requirements-griller— invoke if intent unclear before startingui-quality— token/variant decisions, responsive layout details, and a11y deep-divesanimation-motion— if motion is part of the designthreejs-motion-performance— if 3D / WebGL is involvedsupabase-guardian— if new data shape neededidea-to-prd-tracer— if this is part of a new-feature workflowlinear-operator— for issue creationissue-to-pr— what runs next, per component
Common failure modes
Designer/developer translation breaks — The design has implicit conventions the developer doesn't catch (or vice versa). The skill's job is to make them explicit before code is written. If you find yourself implementing and "guessing" — go back, the spec wasn't done.
Lovable code quality false positive — Lovable output looks polished. It's almost never production-ready. Default disposition is rewrite.
Token sprawl — Every design adds 3 new colors. Three months later you have 60 colors and no system. Push back on token additions. Reuse aggressively. Propose a new token only when no existing one is close enough.
Skipping states — Design typically shows 1-2 states. Production needs 6-8. The state map is what catches this gap.
Implementing before spec is approved — Eager implementation = re-implementing later. Spec → approval → implementation → QA. Don't skip steps.
No QA checklist made upfront — Then preview QA becomes a guess. Always produce the checklist as part of the handoff.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/figma-lovable-handoff/SKILL.md
Popularity
- Repository stars
- 3
- 7-day trend score
- 7
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/amnafarzy/ravie/figma-lovable-handoff)