figma-design-review

AI-powered design review for Figma components with weighted dual-scoring system. Evaluates Style Guide Implementation (70%) and LLM Metadata Accessibility (30%). For export, hands off to atomic-design

SKILL.md

GitHub

Figma Design Review Skill

Score Figma component designs with weighted dual-scoring. This skill is PHASE 1 (score components, identify issues, recommend fixes, determine export readiness). Export is PHASE 2, owned by atomic-design (classify level A/M/O, generate files, extract tokens).

Score >= 70? -> Hand off to atomic-design.

When to Use

Trigger Action
Designer selects component Full design review
"Review this component" Scored report
"What's missing?" Gap analysis
"Is this ready for export?" Export readiness check
Score < 80 Improvement roadmap

Weights

Dimension Weight Focus
Style Guide Implementation 70% Code quality, consistency, accessibility
LLM Metadata Accessibility 30% Documentation for AI code generation

Dimension 1: Style Guide Implementation (70%)

Group (pts) Criterion Points
Variant Structure (25) Complete variant matrix 10
Consistent naming (Size=X, State=Y) 5
TypeScript interface defined 5
Logical organization 5
Token System (25) CSS Variables for colors 10
CSS Variables for typography 5
CSS Variables for spacing 5
Semantic token naming 5
Visual Consistency (20) Consistent font-weight 5
Consistent border-radius 5
Consistent padding/spacing 5
No layout artifacts 5
Accessibility States (20) Focus state present 8
Disabled state 6
Hover state 4
Touch target >= 44px 2
Code Quality (10) No duplication 5
Clean conditionals 3
Proper defaults 2

Dimension 2: LLM Metadata Accessibility (30%)

Group (pts) Criterion Points
Component Description (20) Primary description 10
Use case specified 5
Business context 5
Searchability (15) Tags present 8
Tags comprehensive 7
Development Metadata (25) testId 5
ariaLabel 5
analytics 5
category 5
level 5
Usage Guidelines (20) Do's list 7
Don'ts list 7
Notes 6
Technical Specs (20) Design tokens documented 8
Specs (minWidth, etc.) 6
A11y requirements 6

Score Interpretation

Score Grade Export Status
90-100 Excellent Ready
80-89 Good Ready with notes
70-79 Acceptable Fix critical first
60-69 Needs Work Not ready
< 60 Poor Major rework

Review Framework

Step 1 - Extract Data: name/structure from Figma; code from get_design_context; screenshot from get_screenshot; metadata from description.

Step 2 - Evaluate Implementation (70%): variant count expected vs actual; token usage check; visual consistency; accessibility states; code quality.

Step 3 - Evaluate LLM Accessibility (30%): description present? tags comprehensive? dev metadata complete? usage guidelines? technical specs?

Step 4 - Calculate:

implementation_score = variant + token + visual + a11y + code
llm_score = desc + search + dev + usage + specs
weighted = (implementation * 0.7) + (llm * 0.3)

Step 5 - Generate Report: strengths (max 6); weaknesses (max 5, each with fix); recommendations by priority; export readiness.

Report Output Format

## Component Evaluation: [Name]

### Style Guide Implementation: **[XX]/100**

**Strengths:**
- [Positive finding]

**Weaknesses:**
- [Issue] -> **Fix:** [Solution]

### LLM Metadata: **[XX]/100**

**Includes:**
- [Present metadata]

**Missing:**
- [Missing metadata]

## Final Weighted Score

| Criterion | Score | Weight | Contribution |
|-----------|-------|--------|--------------|
| Implementation | XX | 70% | XX.X |
| LLM Accessibility | XX | 30% | XX.X |
| **Total** | | | **XX.X/100** |

### Recommendations:
1. [Recommendation] - [explanation]

Key Rules

  • Every weakness MUST include a fix
  • Every issue MUST name specific field
  • Strengths capped at 6
  • Weaknesses capped at 5
  • Recommendations ordered by priority
  • English only

References

File Purpose
references/audit-summary-format.md Output format
references/scoring-rubric.md Scoring criteria
references/common-issues.md Frequent problems
../atomic-design/SKILL.md Export skill