Claude-Code-Workflow/.claude/commands/workflow/ui-design/layout-extract.md

name: layout-extract description: Extract structural layout information from reference images, URLs, or text prompts using Claude analysis argument-hint: [--base-path ] [--session ] [--images ""] [--urls ""] [--prompt ""] [--targets ""] [--mode <imitate|explore>] [--variants ] [--device-type <desktop|mobile|tablet|responsive>] allowed-tools: TodoWrite(), Read(), Write(), Glob(), Bash(), Task(ui-design-agent), mcp__exa__web_search_exa()

Layout Extraction Command

Overview

Extract structural layout information from reference images, URLs, or text prompts using AI analysis. This command separates the "scaffolding" (HTML structure and CSS layout) from the "paint" (visual tokens handled by style-extract).

Strategy: AI-Driven Structural Analysis

• Agent-Powered: Uses ui-design-agent for deep structural analysis
• Dual-Mode:
  • imitate: High-fidelity replication of single layout structure
  • explore: Multiple structurally distinct layout variations
• Single Output: layout-templates.json with DOM structure, component hierarchy, and CSS layout rules
• Device-Aware: Optimized for specific device types (desktop, mobile, tablet, responsive)
• Token-Based: CSS uses var() placeholders for spacing and breakpoints

Phase 0: Setup & Input Validation

Step 1: Detect Input, Mode & Targets

# Detect input source
# Priority: --urls + --images → hybrid | --urls → url | --images → image | --prompt → text

# Parse URLs if provided (format: "target:url,target:url,...")
IF --urls:
    url_list = []
    FOR pair IN split(--urls, ","):
        IF ":" IN pair:
            target, url = pair.split(":", 1)
            url_list.append({target: target.strip(), url: url.strip()})
        ELSE:
            # Single URL without target
            url_list.append({target: "page", url: pair.strip()})

    has_urls = true
ELSE:
    has_urls = false
    url_list = []

# Determine extraction mode
extraction_mode = --mode OR "imitate"  # "imitate" or "explore"

# Set variants count based on mode
IF extraction_mode == "imitate":
    variants_count = 1  # Force single variant (ignore --variants)
ELSE IF extraction_mode == "explore":
    variants_count = --variants OR 3  # Default to 3
    VALIDATE: 1 <= variants_count <= 5

# Resolve targets
# Priority: --targets → url_list targets → prompt analysis → default ["page"]
IF --targets:
    targets = split(--targets, ",")
ELSE IF has_urls:
    targets = [url_info.target for url_info in url_list]
ELSE IF --prompt:
    targets = extract_from_prompt(--prompt)
ELSE:
    targets = ["page"]

# Resolve device type
device_type = --device-type OR "responsive"  # desktop|mobile|tablet|responsive

# Determine base path (auto-detect and convert to absolute)
relative_path=$(find .workflow -type d -name "design-run-*" | head -1)
base_path=$(cd "$relative_path" && pwd)
bash(test -d "$base_path" && echo "✓ Base path: $base_path" || echo "✗ Path not found")
# OR use --base-path / --session parameters

Step 2: Load Inputs & Create Directories

# For image mode
bash(ls {images_pattern})  # Expand glob pattern
Read({image_path})  # Load each image

# For URL mode
# Parse URL list format: "target:url,target:url"
# Validate URLs are accessible

# For text mode
# Validate --prompt is non-empty

# Create output directory
bash(mkdir -p {base_path}/layout-extraction)

Step 2.5: Extract DOM Structure (URL Mode - Auto-Trigger)

# AUTO-TRIGGER: If URLs are available (from --urls parameter), automatically extract real DOM structure
# This provides accurate layout data to supplement visual analysis

# Check if URLs provided via --urls parameter
IF --urls AND url_list:
    REPORT: "🔍 Auto-triggering URL mode: Extracting DOM structure"

    bash(mkdir -p {base_path}/.intermediates/layout-analysis)

    # For each URL in url_list:
    FOR url_info IN url_list:
        target = url_info.target
        url = url_info.url

        IF mcp_chrome_devtools_available:
            REPORT: "   Processing: {target} ({url})"

            # Read extraction script
            script_content = Read(~/.claude/scripts/extract-layout-structure.js)

            # Open page in Chrome DevTools
            mcp__chrome-devtools__navigate_page(url=url)

            # Execute layout extraction script
            result = mcp__chrome-devtools__evaluate_script(function=script_content)

            # Save DOM structure for this target (intermediate file)
            Write({base_path}/.intermediates/layout-analysis/dom-structure-{target}.json, result)

            REPORT: "   ✅ DOM structure extracted for '{target}'"
        ELSE:
            REPORT: "   ⚠️ Chrome DevTools MCP not available, falling back to visual analysis"
            BREAK

    dom_structure_available = mcp_chrome_devtools_available
ELSE:
    dom_structure_available = false

Extraction Script Reference: ~/.claude/scripts/extract-layout-structure.js

Usage: Read the script file and use content directly in mcp__chrome-devtools__evaluate_script()

Script returns:

• metadata: Extraction timestamp, URL, method, version
• patterns: Layout pattern statistics (flexColumn, flexRow, grid counts)
• structure: Hierarchical DOM tree with layout properties
• exploration: (Optional) Progressive exploration results when standard selectors fail

Benefits:

• ✅ Real flex/grid configuration (justifyContent, alignItems, gap, etc.)
• ✅ Accurate element bounds (x, y, width, height)
• ✅ Structural hierarchy with depth control
• ✅ Layout pattern identification (flex-row, flex-column, grid-NCol)
• ✅ Progressive exploration: Auto-discovers missing selectors

Progressive Exploration Strategy (v2.2.0+):

When script finds <3 main containers, it automatically:

1. Scans all large visible containers (≥500×300px)
2. Extracts class patterns matching: main|content|wrapper|container|page|layout|app
3. Suggests new selectors to add to script
4. Returns exploration data in result.exploration:

   {
     "triggered": true,
     "discoveredCandidates": [{classes, bounds, display}],
     "suggestedSelectors": [".wrapper", ".page-index"],
     "recommendation": ".wrapper, .page-index, .app-container"
   }
   

Using Exploration Results:

// After extraction, check for suggestions
IF result.exploration?.triggered:
    REPORT: result.exploration.warning
    REPORT: "Suggested selectors: " + result.exploration.recommendation

    // Update script by adding to commonClassSelectors array
    // Then re-run extraction for better coverage

Selector Update Workflow:

1. Run extraction on unfamiliar site
2. Check result.exploration.suggestedSelectors
3. Add relevant selectors to script's commonClassSelectors
4. Re-run extraction → improved container detection

Step 3: Memory Check

# 1. Check if inputs cached in session memory
IF session_has_inputs: SKIP Step 2 file reading

# 2. Check if output already exists
bash(test -f {base_path}/layout-extraction/layout-templates.json && echo "exists")
IF exists: SKIP to completion

---

Phase 0 Output: input_mode, base_path, extraction_mode, variants_count, targets[], device_type, loaded inputs

Phase 1: Layout Concept Generation (Explore Mode Only)

Step 1: Check Extraction Mode

# extraction_mode == "imitate" → skip this phase
# extraction_mode == "explore" → execute this phase

If imitate mode: Skip to Phase 2

Step 2: Gather Layout Inspiration (Explore Mode)

bash(mkdir -p {base_path}/.intermediates/layout-analysis/inspirations)

# For each target: Research via MCP
# mcp__exa__web_search_exa(query="{target} layout patterns {device_type}", numResults=5)

# Write inspiration file
Write({base_path}/.intermediates/layout-analysis/inspirations/{target}-layout-ideas.txt, inspiration_content)

Step 3: Generate Layout Concept Options (Agent Task 1)

Executor: Task(ui-design-agent)

Launch agent to generate variants_count layout concept options for each target:

Task(ui-design-agent): `
  [LAYOUT_CONCEPT_GENERATION_TASK]
  Generate {variants_count} structurally distinct layout concepts for each target

  SESSION: {session_id} | MODE: explore | BASE_PATH: {base_path}
  TARGETS: {targets} | DEVICE_TYPE: {device_type}

  ## Input Analysis
  - Targets: {targets.join(", ")}
  - Device type: {device_type}
  - Layout inspiration: Read inspirations from {base_path}/.intermediates/layout-analysis/inspirations/
  - Visual references: {loaded_images if available}
  ${dom_structure_available ? "- DOM Structure: Read from .intermediates/layout-analysis/dom-structure-*.json" : ""}

  ## Analysis Rules
  - For EACH target, generate {variants_count} structurally DIFFERENT layout concepts
  - Concepts must differ in: grid structure, component arrangement, visual hierarchy
  - Each concept should have distinct navigation pattern, content flow, and responsive behavior

  ## Generate for EACH Target
  For target in {targets}:
    For concept_index in 1..{variants_count}:
      1. **Concept Definition**:
         - concept_name (descriptive, e.g., "Classic Three-Column Holy Grail")
         - design_philosophy (1-2 sentences explaining the structural approach)
         - layout_pattern (e.g., "grid-3col", "flex-row", "single-column", "asymmetric-grid")
         - key_components (array of main layout regions)
         - structural_features (list of distinguishing characteristics)

      2. **Wireframe Preview** (simple text representation):
         - ascii_art (simple ASCII box diagram showing layout structure)
         - Example:
           ┌─────────────────┐
           │     HEADER      │
           ├──┬─────────┬────┤
           │ L│  MAIN   │ R  │
           └──┴─────────┴────┘

  ## Output
  Write single JSON file: {base_path}/.intermediates/layout-analysis/analysis-options.json

  Use schema from INTERACTIVE-DATA-SPEC.md (Layout Extract: analysis-options.json)

  CRITICAL: Use Write() tool immediately after generating complete JSON
`

Step 4: Verify Options File Created

bash(test -f {base_path}/.intermediates/layout-analysis/analysis-options.json && echo "created")

# Quick validation
bash(cat {base_path}/.intermediates/layout-analysis/analysis-options.json | grep -q "layout_concepts" && echo "valid")

Output: analysis-options.json with layout concept options for all targets

---

Phase 1.5: User Confirmation (Explore Mode Only - INTERACTIVE)

Purpose: Allow user to select preferred layout concept(s) for each target before generating detailed templates

Step 1: Load and Present Options

# Read options file
options = Read({base_path}/.intermediates/layout-analysis/analysis-options.json)

# Parse layout concepts
layout_concepts = options.layout_concepts

Step 2: Present Options to User (Per Target)

For each target, present layout concept options and capture selection:

📋 Layout Concept Options for Target: {target}

We've generated {variants_count} structurally different layout concepts for review.
Please select your preferred concept for this target.

{FOR each concept in layout_concepts[target]:
  ═══════════════════════════════════════════════════
  Concept {concept.index}: {concept.concept_name}
  ═══════════════════════════════════════════════════

  Philosophy: {concept.design_philosophy}
  Pattern: {concept.layout_pattern}

  Components:
  {FOR each component in concept.key_components:
    • {component}
  }

  Features:
  {FOR each feature in concept.structural_features:
    • {feature}
  }

  Wireframe:
  {concept.wireframe_preview.ascii_art}

  ═══════════════════════════════════════════════════
}

Step 3: Capture User Selection (Per Target)

// Use AskUserQuestion tool for each target
FOR each target:
  AskUserQuestion({
    questions: [{
      question: "Which layout concept do you prefer for '{target}'?",
      header: "Layout for " + target,
      multiSelect: false,
      options: [
        {FOR each concept in layout_concepts[target]:
          label: "Concept {concept.index}: {concept.concept_name}",
          description: "{concept.design_philosophy}"
        }
      ]
    }]
  })

  // Parse user response
  selected_option_text = user_answer

  // Check for user cancellation
  IF selected_option_text == null OR selected_option_text == "":
      REPORT: "⚠️ User canceled selection. Workflow terminated."
      EXIT workflow

  // Extract concept index from response format "Concept N: Name"
  match = selected_option_text.match(/Concept (\d+):/)
  IF match:
      selected_index = parseInt(match[1])
  ELSE:
      ERROR: "Invalid selection format. Expected 'Concept N: ...' format"
      EXIT workflow

  // Store selection for this target
  selections[target] = {
    selected_index: selected_index,
    concept_name: layout_concepts[target][selected_index-1].concept_name
  }

Step 4: Write User Selection File

# Create user selection JSON
selection_data = {
  "metadata": {
    "selected_at": "{current_timestamp}",
    "selection_type": "per_target",
    "session_id": "{session_id}"
  },
  "selections": selections  // {target: {selected_index, concept_name}}
}

# Write to file
bash(echo '{selection_data}' > {base_path}/.intermediates/layout-analysis/user-selection.json)

# Verify
bash(test -f {base_path}/.intermediates/layout-analysis/user-selection.json && echo "saved")

Step 5: Confirmation Message

✅ Selections recorded!

{FOR each target, selection in selections:
  • {target}: Concept {selection.selected_index} - {selection.concept_name}
}

Proceeding to generate detailed layout templates based on your selections...

Output: user-selection.json with user's choices for all targets

Phase 2: Layout Template Generation (Agent Task 2)

Executor: Task(ui-design-agent) for selected concept(s)

Step 1: Load User Selection (Explore Mode)

# For explore mode, read user selection
IF extraction_mode == "explore":
    selection = Read({base_path}/.intermediates/layout-analysis/user-selection.json)
    selections_per_target = selection.selections

    # Also read the selected concept details from options
    options = Read({base_path}/.intermediates/layout-analysis/analysis-options.json)
    layout_concepts = options.layout_concepts

    # Build selected concepts map
    selected_concepts = {}
    FOR each target in targets:
        selected_index = selections_per_target[target].selected_index
        selected_concepts[target] = layout_concepts[target][selected_index-1]  # 0-indexed
ELSE:
    # Imitate mode - no selection needed
    selected_concepts = null

Step 2: Launch Agent Task

Generate layout templates for selected concepts:

Task(ui-design-agent): `
  [LAYOUT_TEMPLATE_GENERATION_TASK]
  Generate detailed layout templates based on user-selected concepts.
  Focus ONLY on structure and layout. DO NOT concern with visual style (colors, fonts, etc.).

  SESSION: {session_id} | MODE: {extraction_mode} | BASE_PATH: {base_path}
  DEVICE_TYPE: {device_type}

  ${extraction_mode == "explore" ? `
  USER SELECTIONS:
  ${targets.map(target => `
  Target: ${target}
  - Selected Concept: ${selected_concepts[target].concept_name}
  - Philosophy: ${selected_concepts[target].design_philosophy}
  - Pattern: ${selected_concepts[target].layout_pattern}
  - Key Components: ${selected_concepts[target].key_components.join(", ")}
  - Structural Features: ${selected_concepts[target].structural_features.join(", ")}
  `).join("\n")}
  ` : `
  MODE: Imitate - High-fidelity replication of reference layout structure
  TARGETS: ${targets.join(", ")}
  `}

  ## Input Analysis
  - Targets: {targets.join(", ")}
  - Device type: {device_type}
  - Visual references: {loaded_images if available}
  ${dom_structure_available ? "- DOM Structure Data: Read from .intermediates/layout-analysis/dom-structure-*.json - USE THIS for accurate layout properties" : ""}

  ## Generation Rules
  ${extraction_mode == "explore" ? `
  - **Explore Mode**: Develop each user-selected layout concept into a detailed template
  - Use the selected concept's key_components as foundation
  - Apply the selected layout_pattern (grid-3col, flex-row, etc.)
  - Honor the structural_features defined in the concept
  - Expand the concept with complete DOM structure and CSS layout rules
  ` : `
  - **Imitate Mode**: High-fidelity replication of reference layout structure
  ${dom_structure_available ? "- Use DOM structure data as ground truth" : "- Use visual inference"}
  `}
  ${dom_structure_available ? `
  - IMPORTANT: You have access to real DOM structure data with accurate flex/grid properties
  - Use DOM data as primary source for layout properties
  - Extract real flex/grid configurations (display, flexDirection, justifyContent, alignItems, gap)
  - Use actual element bounds for responsive breakpoint decisions
  - Preserve identified patterns from DOM structure
  ` : ""}

  ## Generate for EACH Target
  For target in {targets}:
    ${extraction_mode == "explore" ? "Based on user-selected concept:" : "Based on reference:"}

    1. **DOM Structure**:
       - Semantic HTML5 tags: <header>, <nav>, <main>, <aside>, <section>, <footer>
       - ARIA roles and accessibility attributes
       ${extraction_mode == "explore" ? "- Use key_components from selected concept" : ""}
       ${dom_structure_available ? "- Base on extracted DOM tree from .intermediates" : "- Infer from visual analysis"}
       - Device-specific optimizations for {device_type}

    2. **Component Hierarchy**:
       - Array of main layout regions
       ${extraction_mode == "explore" ? "- Derived from selected concept's key_components" : ""}

    3. **CSS Layout Rules**:
       ${extraction_mode == "explore" ? "- Implement selected layout_pattern" : ""}
       ${dom_structure_available ? "- Use real layout properties from DOM structure data" : "- Focus on Grid, Flexbox, position, alignment"}
       - Use CSS Custom Properties: var(--spacing-*), var(--breakpoint-*)
       - Device-specific styles (mobile-first @media for responsive)
       - NO colors, NO fonts, NO shadows - layout structure only

  ## Output Format
  Write complete layout-templates.json with layout_templates array.
  Each template must include:
  - target (string)
  - variant_id: "layout-1" (always 1 since only selected concept is generated)
  - source_image_path (string): Reference image path
  - device_type (string)
  - design_philosophy (string ${extraction_mode == "explore" ? "- from selected concept" : ""})
  - dom_structure (JSON object)
  - component_hierarchy (array of strings)
  - css_layout_rules (string)

  ## Critical Requirements
  - ✅ Use Write() tool for layout-templates.json
  - ✅ One template per target (only selected concept)
  - ✅ Structure only, no visual styling
  - ✅ Token-based CSS (var())
  ${extraction_mode == "explore" ? "- ✅ Maintain consistency with selected concepts" : ""}
`

Output: Agent generates layout-templates.json with one template per target

Step 2: Write Output File

# Take JSON output from agent
bash(echo '{agent_json_output}' > {base_path}/layout-extraction/layout-templates.json)

# Verify output
bash(test -f {base_path}/layout-extraction/layout-templates.json && echo "exists")
bash(cat {base_path}/layout-extraction/layout-templates.json | grep -q "layout_templates" && echo "valid")

Output: layout-templates.json created and verified

Completion

Todo Update

TodoWrite({todos: [
  {content: "Setup and input validation", status: "completed", activeForm: "Validating inputs"},
  {content: "Layout research (explore mode)", status: "completed", activeForm: "Researching layout patterns"},
  {content: "Layout analysis and synthesis (agent)", status: "completed", activeForm: "Generating layout templates"},
  {content: "Write layout-templates.json", status: "completed", activeForm: "Saving templates"}
]});

Output Message

✅ Layout extraction complete!

Configuration:
- Session: {session_id}
- Extraction Mode: {extraction_mode} (imitate/explore)
- Device Type: {device_type}
- Targets: {targets}
- Variants per Target: {variants_count}
- Total Templates: {targets.length × variants_count}
{IF has_urls AND dom_structure_available:
- 🔍 URL Mode: DOM structure extracted from {len(url_list)} URL(s)
- Accuracy: Real flex/grid properties from live pages
}
{IF has_urls AND NOT dom_structure_available:
- ⚠️ URL Mode: Chrome DevTools unavailable, used visual analysis fallback
}

{IF extraction_mode == "explore":
Layout Research:
- {targets.length} inspiration files generated
- Pattern search focused on {device_type} layouts
}

Generated Templates:
{FOR each template: - Target: {template.target} | Variant: {template.variant_id} | Philosophy: {template.design_philosophy}}

Output File:
- {base_path}/layout-extraction/layout-templates.json
{IF dom_structure_available:
- {base_path}/.intermediates/layout-analysis/dom-structure-*.json ({len(url_list)} files)
}

Next: /workflow:ui-design:generate will combine these structural templates with style systems to produce final prototypes.

Simple Bash Commands

Path Operations

# Find design directory
bash(find .workflow -type d -name "design-run-*" | head -1)

# Create output directories
bash(mkdir -p {base_path}/layout-extraction)
bash(mkdir -p {base_path}/.intermediates/layout-analysis/inspirations)  # explore mode only

Validation Commands

# Check if already extracted
bash(test -f {base_path}/layout-extraction/layout-templates.json && echo "exists")

# Validate JSON structure
bash(cat layout-templates.json | grep -q "layout_templates" && echo "valid")

# Count templates
bash(cat layout-templates.json | grep -c "\"target\":")

File Operations

# Load image references
bash(ls {images_pattern})
Read({image_path})

# Write inspiration files (explore mode)
Write({base_path}/.intermediates/layout-analysis/inspirations/{target}-layout-ideas.txt, content)

# Write layout templates
bash(echo '{json}' > {base_path}/layout-extraction/layout-templates.json)

Output Structure

{base_path}/
├── .intermediates/                    # Intermediate analysis files
│   └── layout-analysis/
│       ├── dom-structure-{target}.json   # Extracted DOM structure (URL mode only)
│       └── inspirations/                 # Explore mode only
│           └── {target}-layout-ideas.txt # Layout inspiration research
└── layout-extraction/                 # Final layout templates
    ├── layout-templates.json          # Structural layout templates
    └── layout-space-analysis.json     # Layout directions (explore mode only)

layout-templates.json Format

{
  "extraction_metadata": {
    "session_id": "...",
    "input_mode": "image|url|prompt|hybrid",
    "extraction_mode": "imitate|explore",
    "device_type": "desktop|mobile|tablet|responsive",
    "timestamp": "...",
    "variants_count": 3,
    "targets": ["home", "dashboard"]
  },
  "layout_templates": [
    {
      "target": "home",
      "variant_id": "layout-1",
      "source_image_path": "{base_path}/screenshots/home.png",
      "device_type": "responsive",
      "design_philosophy": "Responsive 3-column holy grail layout with fixed header and footer",
      "dom_structure": {
        "tag": "body",
        "children": [
          {
            "tag": "header",
            "attributes": {"class": "layout-header"},
            "children": [{"tag": "nav"}]
          },
          {
            "tag": "div",
            "attributes": {"class": "layout-main-wrapper"},
            "children": [
              {"tag": "main", "attributes": {"class": "layout-main-content"}},
              {"tag": "aside", "attributes": {"class": "layout-sidebar-left"}},
              {"tag": "aside", "attributes": {"class": "layout-sidebar-right"}}
            ]
          },
          {"tag": "footer", "attributes": {"class": "layout-footer"}}
        ]
      },
      "component_hierarchy": [
        "header",
        "main-content",
        "sidebar-left",
        "sidebar-right",
        "footer"
      ],
      "css_layout_rules": ".layout-main-wrapper { display: grid; grid-template-columns: 1fr 3fr 1fr; gap: var(--spacing-6); } @media (max-width: var(--breakpoint-md)) { .layout-main-wrapper { grid-template-columns: 1fr; } }"
    }
  ]
}

Requirements: Token-based CSS (var()), semantic HTML5, device-specific structure, accessibility attributes

Error Handling

Common Errors

ERROR: No inputs provided
→ Provide --images, --urls, or --prompt

ERROR: Invalid target name
→ Use lowercase, alphanumeric, hyphens only

ERROR: Agent task failed
→ Check agent output, retry with simplified prompt

ERROR: MCP search failed (explore mode)
→ Check network, retry

Recovery Strategies

• Partial success: Keep successfully extracted templates
• Invalid JSON: Retry with stricter format requirements
• Missing inspiration: Works without (less informed exploration)

Key Features

• Auto-Trigger URL Mode - Automatically extracts DOM structure when --urls provided (no manual flag needed)
• Hybrid Extraction Strategy - Combines real DOM structure data with AI visual analysis
• Accurate Layout Properties - Chrome DevTools extracts real flex/grid configurations, bounds, and hierarchy
• Separation of Concerns - Decouples layout (structure) from style (visuals)
• Structural Exploration - Explore mode enables A/B testing of different layouts
• Token-Based Layout - CSS uses var() placeholders for instant design system adaptation
• Device-Specific - Tailored structures for different screen sizes
• Graceful Fallback - Falls back to visual analysis if Chrome DevTools unavailable
• Foundation for Assembly - Provides structural blueprint for refactored generate command
• Agent-Powered - Deep structural analysis with AI

Integration

Workflow Position: Between style extraction and prototype generation

New Workflow:

1. /workflow:ui-design:style-extract → design-tokens.json + style-guide.md (Complete design systems)
2. /workflow:ui-design:layout-extract → layout-templates.json (Structural templates)
3. /workflow:ui-design:generate (Pure assembler):
   • Reads: design-tokens.json + layout-templates.json
   • Action: For each style × layout combination:
     1. Build HTML from dom_structure
     2. Create layout CSS from css_layout_rules
     3. Link design tokens CSS
     4. Inject placeholder content
   • Output: Complete token-driven HTML/CSS prototypes

Input: Reference images, URLs, or text prompts Output: layout-templates.json for /workflow:ui-design:generate Next: /workflow:ui-design:generate --session {session_id}