browser-testing

name: browser-testing description: > Browser-based testing with intelligent tool selection across chrome-cdp, Chrome extension, Playwright, and PinchTab. Use when testing UIs, verifying visual output, running end-to-end flows, checking network/console, or performing accessibility audits in a browser. Triggers on: "test in browser", "browser test", "visual test", "check the UI", "screenshot test", "end-to-end test", "verify the page", "check accessibility".

Browser Testing Skill

Tool Selection

Pick the right tool based on what you're testing:

Default choice: chrome-cdp for quick inspection of live tabs (lightest, no extension). Chrome extension for visual/coordinate-based interaction. Playwright for headless or scripted. PinchTab for reading content or parallel tasks.

Tool Reference

chrome-cdp (CLI — ~/.claude/skills/chrome-cdp/scripts/cdp.mjs)

Direct CDP via WebSocket. No extension, no Puppeteer. Requires Chrome remote debugging enabled (chrome://inspect/#remote-debugging) and Node.js 22+.

cdp.mjs list                        # List open tabs (shows targetId prefixes)
cdp.mjs shot   <target> [file]      # Screenshot viewport → /tmp/screenshot.png
cdp.mjs snap   <target>             # Accessibility tree (compact, semantic)
cdp.mjs html   <target> [selector]  # Full HTML or scoped to CSS selector
cdp.mjs eval   <target> "expr"      # Evaluate JS in page context
cdp.mjs nav    <target> <url>       # Navigate and wait for load
cdp.mjs net    <target>             # Network resource timing
cdp.mjs click  <target> "selector"  # Click element by CSS selector
cdp.mjs clickxy <target> <x> <y>    # Click at CSS pixel coordinates
cdp.mjs type   <target> "text"      # Type at focused element (works in cross-origin iframes)
cdp.mjs stop   [target]             # Stop daemon(s)

<target> is a unique prefix of the targetId from list. Daemons auto-exit after 20 min idle.

Chrome Extension (mcp__claude-in-chrome__*)

tabs_context_mcp     → Get current browser tabs (ALWAYS call first)
tabs_create_mcp      → Open new tab
navigate             → Go to URL
read_page            → Get page content
get_page_text        → Get text-only content
find                 → Search for text on page
computer             → Click, scroll, interact at coordinates
form_input           → Fill form fields
read_console_messages → Read console output (use pattern param to filter)
read_network_requests → Inspect network traffic
javascript_tool      → Execute JS in page context

Playwright (mcp__plugin_playwright_playwright__*)

browser_navigate     → Go to URL
browser_snapshot     → Get accessibility tree snapshot
browser_take_screenshot → Capture screenshot
browser_click        → Click element by ref or text
browser_fill_form    → Fill form fields
browser_type         → Type text
browser_press_key    → Keyboard input
browser_evaluate     → Run JS in page
browser_console_messages → Read console
browser_network_requests → Inspect network
browser_resize       → Set viewport size
browser_tabs         → List/switch tabs

PinchTab (CLI or HTTP API)

pinchtab                    # Start server (port 9867)
pinchtab nav <url>          # Navigate
pinchtab snap -i -c         # Get page snapshot (interactive + clickable refs)
pinchtab click <ref>        # Click element by ref (e.g., e5)
pinchtab type <ref> "text"  # Type into element
pinchtab text               # Get page text (token-efficient)
pinchtab eval "js code"     # Execute JavaScript
pinchtab screenshot         # Capture screenshot
pinchtab tabs               # List tabs

HTTP API (for programmatic use):

curl http://localhost:9867/api/v1/navigate -d '{"url":"https://example.com"}'
curl http://localhost:9867/api/v1/snapshot
curl http://localhost:9867/api/v1/click -d '{"ref":"e5"}'

Standard Testing Workflow

1. Pre-flight

• Ensure local server is running (check package.json scripts, Procfile, or start manually)
• Choose tool based on selection matrix above
• If using Chrome: call tabs_context_mcp first to see open tabs

2. Navigate & Baseline

• Open the target URL
• Capture initial state (screenshot or snapshot)
• Check console for errors (pattern param to filter noise)

3. Test Interactions

• Click through the user flow
• Fill forms with test data
• Verify state changes after each action
• Capture screenshots at key checkpoints

4. Verify

• Compare final state against expectations
• Check network requests for expected API calls
• Verify no console errors
• Test error states (invalid input, network failure)

5. Report

• Summarize pass/fail with evidence
• Include screenshots for visual verification
• Note any console errors or failed network requests

Accessibility Testing Checklist

Run these checks on every UI test:

• Keyboard navigation: Tab through all interactive elements. Focus order is logical.
• Focus visibility: Focus ring is visible on all interactive elements.
• Screen reader landmarks: Page has <main>, <nav>, <header>, <footer> as appropriate.
• ARIA labels: All interactive elements have accessible names.
• Color contrast: Text meets WCAG AA (4.5:1 normal text, 3:1 large text).
• Form labels: Every input has an associated <label>.
• Alt text: Images have descriptive alt text (or alt="" for decorative).
• Error messages: Form errors are announced and associated with fields.

Use chrome-cdp snap for quick accessibility tree dumps (lightest option), or PinchTab snap -i -c for structured element refs.

Common Patterns

Screenshot Comparison

1. Navigate to page
2. Take screenshot → save as baseline
3. Make change (or interact)
4. Take screenshot → save as result
5. Compare visually or report differences

Network Verification

1. Start monitoring network requests
2. Perform action (form submit, page load, etc.)
3. Read network requests
4. Verify expected endpoints were called with correct payloads
5. Check response status codes

Mobile Responsive Testing

1. Use Playwright browser_resize to set viewport:
   - Mobile: 375x812 (iPhone)
   - Tablet: 768x1024 (iPad)
   - Desktop: 1440x900
2. Take screenshot at each breakpoint
3. Verify layout adapts correctly