verify-ui

verify-ui — Browser-Based UI Verification

Visually verify UI changes using agent-browser: screenshot + accessibility check + error scan. Add deep for interactive testing.

Usage

/verify-ui                      # Quick mode: screenshot + a11y + errors
/verify-ui /dashboard           # Quick mode at specific path
/verify-ui deep                 # Deep mode: also clicks, fills, navigates
/verify-ui /dashboard deep      # Deep mode at specific path

Step 1: Detect Project & Port

Read package.json scripts to find the dev command and port:

cat package.json | grep -E '"dev"|"start"'

Framework port defaults:

Framework	Default Port
Astro	4321
Vite	5173
Next.js	3000
Generic	3000

Check --port flags in dev scripts first, fall back to framework defaults.

Step 2: Check if Server is Running

lsof -i :<port> | grep LISTEN

• Running → skip to Step 3
• Not running → detect package manager and start:

# Detect package manager from lockfile
ls bun.lock 2>/dev/null && PM=bun || \
ls pnpm-lock.yaml 2>/dev/null && PM=pnpm || \
ls yarn.lock 2>/dev/null && PM=yarn || PM=npm

# Start in background
$PM run dev &

Then poll until ready (up to 30s):

agent-browser open http://localhost:<port>

Track whether you started the server so you can offer cleanup at the end.

Step 3: Navigate

# Default: project root
agent-browser open http://localhost:<port>

# With path argument
agent-browser open http://localhost:<port>/the-path

# With full URL
agent-browser open <url>

Wait for page load:

agent-browser wait --load networkidle

Step 4: Quick Verification (always run)

Run these three checks in sequence:

4a. Screenshot

agent-browser screenshot

Analyze the screenshot: layout issues, broken styles, missing content, visual regressions.

4b. Accessibility snapshot

agent-browser snapshot -i

Check for: interactive elements present, proper labels, navigation structure, no missing headings.

4c. Console errors

agent-browser errors

Flag any JS errors, failed network requests, or console warnings.

4d. Report Summarize inline:

• Screenshot: what you see (layout, content, styles)
• Elements found: key interactive components
• Errors: any JS errors (NONE is good)
• Issues: anything that looks broken

Step 5: Deep Verification (only when deep argument passed)

All of quick mode, plus:

5a. Identify interactive elements From the snapshot refs (@e1, @e2, etc.), identify:

• Primary CTA buttons → click them, screenshot after
• Navigation links → click, verify URL changes, go back
• Forms → fill with test data, submit, check validation

5b. Click primary buttons

agent-browser click @e<ref>
agent-browser wait --load networkidle
agent-browser screenshot
agent-browser errors

5c. Test forms

agent-browser fill @e<ref> "test@example.com"  # email fields
agent-browser fill @e<ref> "Test Input"          # text fields
agent-browser click @e<submit-ref>
agent-browser wait --load networkidle
agent-browser screenshot

5d. Check responsive behavior

agent-browser set viewport 375 812   # Mobile
agent-browser screenshot

agent-browser set viewport 768 1024  # Tablet
agent-browser screenshot

agent-browser set viewport 1440 900  # Desktop
agent-browser screenshot

5e. Report Include interaction outcomes, navigation flow, form behavior, and responsive screenshots.

Step 6: Cleanup

If you started the dev server:

kill %1  # or the PID you tracked

Always close the browser session:

agent-browser close

Common Issues

Symptom	Fix
Port already in use	Check lsof -i :<port> — something else may be running there
agent-browser open times out	Server may need more time to start; wait and retry
Screenshot is blank	Wait for networkidle before screenshotting
No interactive elements in snapshot	Page may be fully static; that's OK — report what you see
Server won't start	Check package.json dev script, report the error