visual-testing（视觉测试）

Visual Testing Skill

Perform visual testing, UI verification, and design comparison using screenshots, responsive testing, and Figma integration.

When to Use

This skill activates when:

• User wants to verify UI appearance
• User asks to compare page with design or Figma
• User needs responsive design testing
• User wants to check for visual regressions
• User mentions UI bugs or styling issues
• User asks about design parity
• User wants to validate implementation against design

Capabilities

Screenshot Capture

browser-devtools-cli content take-screenshot --name "homepage"
browser-devtools-cli content take-screenshot --name "element" --selector ".hero"
browser-devtools-cli content take-screenshot --name "fullpage" --full-page
browser-devtools-cli content take-screenshot --name "photo" --type jpeg --quality 80
browser-devtools-cli content save-as-pdf --name "page" --format A4

Annotated Screenshot (ref labels)

Overlay numbered labels [1],[2] on interactive elements. Call a11y take-aria-snapshot first.

browser-devtools-cli a11y take-aria-snapshot
browser-devtools-cli content take-screenshot --annotate --name "annotated"
browser-devtools-cli content take-screenshot --annotate --annotate-content --name "with-headings"

Responsive Testing

browser-devtools-cli interaction resize-viewport --width 375 --height 667   # Mobile
browser-devtools-cli interaction resize-viewport --width 768 --height 1024  # Tablet
browser-devtools-cli interaction resize-viewport --width 1920 --height 1080 # Desktop

Figma Design Comparison

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123xyz" \
  --figma-node-id "12:34"

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123xyz" \
  --figma-node-id "12:34" \
  --selector ".hero-section"

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123xyz" \
  --figma-node-id "12:34" \
  --mssim-mode semantic

DOM Inspection

browser-devtools-cli content get-as-html
browser-devtools-cli content get-as-html --selector ".component"
browser-devtools-cli content get-as-text --selector ".content"

Viewport Presets

Prerequisites for Figma Comparison

Figma API Access:

• Set FIGMA_ACCESS_TOKEN environment variable
• Get token from Figma account settings

Optional (for enhanced comparison):

• AWS credentials for image/text embeddings
• Cloud inference profile configuration

Visual Testing Workflow

SESSION="--session-id visual-test"

# 1. Navigate
browser-devtools-cli $SESSION navigation go-to --url "https://example.com"
browser-devtools-cli $SESSION sync wait-for-network-idle

# 2. Capture desktop
browser-devtools-cli $SESSION content take-screenshot --name "desktop"

# 3. Test mobile viewport
browser-devtools-cli $SESSION interaction resize-viewport --width 375 --height 667
browser-devtools-cli $SESSION content take-screenshot --name "mobile"

# 4. Test tablet viewport
browser-devtools-cli $SESSION interaction resize-viewport --width 768 --height 1024
browser-devtools-cli $SESSION content take-screenshot --name "tablet"

# 5. Cleanup
browser-devtools-cli session delete visual-test

Figma Comparison Workflow

1. Get Figma Reference

Find the Figma file key and node ID:

• File key: From URL figma.com/file/{FILE_KEY}/...
• Node ID: Select frame → Copy link → Extract node-id parameter

2. Compare with Live Page

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123xyz" \
  --figma-node-id "12:34" \
  --selector ".hero-section"

3. Analyze Results

• Score 0.9-1.0: Excellent match
• Score 0.7-0.9: Minor differences
• Score 0.5-0.7: Significant differences
• Score < 0.5: Major mismatch

4. Investigate Differences

If score is low:

# Take screenshot of the area
browser-devtools-cli content take-screenshot --name "mismatch" --selector ".hero-section"

# Check HTML structure
browser-devtools-cli content get-as-html --selector ".hero-section"

Comparison Modes

Semantic Mode (Default)

• More tolerant of text/data differences
• Focuses on layout and structure
• Good for comparing sample vs real data

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123" \
  --figma-node-id "1:2" \
  --mssim-mode semantic

Raw Mode

• Stricter pixel comparison
• Expects near-identical output
• Good for static content

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123" \
  --figma-node-id "1:2" \
  --mssim-mode raw

Common Scenarios

Full Page Comparison

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123" \
  --figma-node-id "1:2" \
  --full-page

Component Comparison

browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123" \
  --figma-node-id "5:10" \
  --selector ".card-component"

Responsive Testing with Figma

# Resize viewport first
browser-devtools-cli interaction resize-viewport --width 375 --height 667

# Then compare with mobile design
browser-devtools-cli figma compare-page-with-design \
  --figma-file-key "ABC123" \
  --figma-node-id "20:30"

Common Visual Checks

• Element visibility at breakpoints
• Text overflow and truncation
• Image aspect ratios
• Color and typography consistency
• Spacing and alignment
• Interactive state styling

Best Practices

1. Use semantic mode for real data vs sample data
2. Compare components not just full pages
3. Test multiple viewports for responsive designs
4. Document Figma references for CI/CD
5. Set baseline scores for regression testing
6. Review notes in response for signal details
7. Take screenshots at key states (initial, hover, active)