/analyze-website

claude-frontforge analyze-website

Extract design systems from live websites through vision-based analysis.

Usage

/analyze-website <url>                    # Analyze single site
/analyze-website <url> --focus=colors     # Focus on specific tokens
/analyze-website <url1> <url2> --compare  # Compare multiple sites

Options

• --focus=<area>: typography, colors, spacing, layout, surfaces, all (default: all)
• --format=<format>: json, css, tailwind, figma (default: json)
• --compare: Enable comparison mode for multiple URLs

What Gets Extracted

Typography

• Font families (classified by type: sans, serif, mono, display)
• Size scale with relationships
• Weight usage patterns
• Line height conventions
• Confidence: 80-85%

Colors

• Primary/secondary/accent colors
• Neutral scale
• Semantic colors (success, warning, error)
• Confidence: 85-90% (±10-15% hex variance)

Spacing

• Base unit detection (4px, 8px, 10px)
• Scale pattern identification
• Component-specific spacing
• Confidence: 75-80%

Layout

• Grid system (columns, gutters)
• Container widths
• Section spacing
• Confidence: 70-80%

Surfaces

• Shadow patterns and depths
• Border treatments
• Backdrop effects
• Confidence: 80-85%

Analysis Protocol

3-Pass Capture

1. Structure Pass (1440px desktop)

   • Major sections identification
   • Layout grid analysis
   • Visual hierarchy mapping

2. Component Pass (Element inspection)

   • Button patterns
   • Card treatments
   • Input styles
   • Navigation patterns

3. Interaction Pass (States + responsive)

   • Hover states
   • Focus indicators
   • Breakpoint behaviors (768px, 375px)

Output Example

{
  "metadata": {
    "url": "stripe.com",
    "analyzedAt": "2026-02-02T10:30:00Z",
    "overallConfidence": "82%"
  },
  "tokens": {
    "typography": {
      "families": {
        "primary": "Inter (sans-serif)",
        "mono": "JetBrains Mono"
      },
      "scale": ["12px", "14px", "16px", "18px", "24px", "32px", "48px"],
      "weights": ["400", "500", "600", "700"],
      "confidence": "85%"
    },
    "colors": {
      "primary": "#635BFF",
      "secondary": "#0A2540",
      "neutral": ["#F6F9FC", "#E3E8EE", "#8792A2", "#425466", "#0A2540"],
      "semantic": {
        "success": "#30D158",
        "warning": "#FF9F0A",
        "error": "#FF453A"
      },
      "confidence": "88%"
    },
    "spacing": {
      "base": "4px",
      "scale": ["4px", "8px", "16px", "24px", "32px", "48px", "64px"],
      "confidence": "78%"
    },
    "surfaces": {
      "shadows": [
        "0 1px 3px rgba(0,0,0,0.08)",
        "0 4px 12px rgba(0,0,0,0.1)",
        "0 12px 24px rgba(0,0,0,0.12)"
      ],
      "borders": "1px solid rgba(0,0,0,0.08)",
      "radius": ["4px", "8px", "12px"],
      "confidence": "82%"
    }
  },
  "contentAnalysis": {
    "siteType": "SaaS / Fintech",
    "voiceTone": "Professional, confident, technical",
    "industrySignals": ["payments", "developer-focused", "enterprise"],
    "confidence": "92%"
  },
  "strategyInference": {
    "primaryGoal": "Developer trust and enterprise credibility",
    "audienceSignals": ["Technical decision-makers", "Developers"],
    "designTradeoffs": [
      "Chose density over breathing room",
      "Prioritized information over delight"
    ]
  },
  "suggestedDirection": "Sophistication & Trust",
  "decisionTraces": [
    "Cool color palette (blue/slate) → Professional, trustworthy",
    "Layered shadows → Premium, dimensional feel",
    "Monospace in code blocks → Developer credibility"
  ]
}

Comparison Mode

When using --compare with multiple URLs:

/analyze-website stripe.com linear.app mercury.com --compare

Outputs a comparison matrix:

| Aspect | Stripe | Linear | Mercury |
|--------|--------|--------|---------|
| Direction | Sophistication | Precision | Sophistication |
| Base spacing | 4px | 4px | 8px |
| Depth strategy | Layered | Borders | Subtle shadows |
| Typography | Inter | Inter | SF Pro |
| Primary color | #635BFF | #5E6AD2 | #2563EB |

Known Limitations

Be transparent about accuracy:

• Font family identification: ~60% accurate (classify by type anatomy)
• Hex values: ±10-15% variance from screenshots
• Cannot detect: Exact transition timings, CSS implementation details
• Responsive: Requires separate captures per viewport

Integration with Frontforge

After analysis, you can:

1. Create system.md: /init with extracted tokens
2. Generate tokens: /generate-tokens --format=tailwind
3. Apply to project: Use extracted patterns as foundation

Example Workflow

User: /analyze-website linear.app

Claude: Analyzing linear.app...

📊 Analysis Complete (Confidence: 85%)

**Direction Detected:** Precision & Density
**Key Characteristics:**
- Tight 4px grid
- Borders-only depth (minimal shadows)
- Monochromatic palette with purple accent
- Dense information display

**Typography:**
- Primary: Inter (sans-serif)
- Scale: 11, 12, 13, 14, 16, 20, 24px
- Weights: 400, 500, 600

**Colors:**
- Primary: #5E6AD2 (purple)
- Background: #1A1A1F (dark)
- Text: #E8E8ED (light)

**Strategy Inference:**
Linear optimizes for power users who need information density.
The minimal depth (borders-only) keeps focus on content.

Create .frontforge/system.md with these patterns? (y/n)