Visual QA

Visual QA Checkpoint

Visual quality assurance checkpoint that uses browser automation to verify the UI.

What This Skill Does

1. Checks browser console for errors
2. Takes screenshot of current page
3. Analyzes for obvious visual issues
4. Writes result to checkpoint file

Prerequisites

• TabzChrome extension running
• Application loaded in browser tab
• tabz MCP server connected

Workflow

Step 1: Check Console Errors

# Get console logs filtered for errors
mcp-cli info tabz/tabz_get_console_logs
mcp-cli call tabz/tabz_get_console_logs '{"level": "error"}'

Important: Console errors don't automatically fail - evaluate if they're blocking:

• JS runtime errors that prevent functionality = FAIL
• 404s for optional resources = WARNING
• Deprecation warnings = INFO

Step 2: Take Screenshot

# Capture current page state
mcp-cli info tabz/tabz_screenshot
mcp-cli call tabz/tabz_screenshot '{}'

The screenshot file path is returned. Read it to visually inspect the page.

Step 3: Check Page State

# Verify page loaded correctly
mcp-cli info tabz/tabz_get_page_info
mcp-cli call tabz/tabz_get_page_info '{}'

Verify:

• Page title is not error page
• URL matches expected
• Page is not stuck loading

Step 4: Optional - Check Network Errors

If functionality seems broken:

# Enable capture first (if not already)
mcp-cli call tabz/tabz_enable_network_capture '{}'

# Trigger the problematic action, then:
mcp-cli call tabz/tabz_get_network_requests '{"statusMin": 400}'

Step 5: Parse and Write Result

Create structured result:

{
  "checkpoint": "visual-qa",
  "timestamp": "2026-01-19T12:00:00Z",
  "passed": true,
  "issues": [],
  "screenshots": ["/path/to/screenshot.png"],
  "console_errors": 0,
  "summary": "Page loads correctly, no visual issues detected"
}

Result Fields:

• passed: true if no blocking visual/console issues
• issues: array of {severity: "error"|"warning"|"info", message: string, type: "console"|"visual"|"network"}
• screenshots: array of screenshot file paths
• console_errors: count of console errors found
• summary: brief human-readable summary

Step 6: Write Checkpoint File

mkdir -p .checkpoints
cat > .checkpoints/visual-qa.json << 'EOF'
{
  "checkpoint": "visual-qa",
  "timestamp": "...",
  "passed": true,
  "issues": [],
  "screenshots": [...],
  "console_errors": 0,
  "summary": "..."
}
EOF

Decision Criteria

Pass if:

• Page loads without critical console errors
• No obvious visual breakage (blank page, missing components)
• Key functionality appears present

Fail if:

• JS errors prevent page from rendering
• Page shows error state or blank
• Critical UI elements missing
• API calls failing (5xx errors)

Warning (pass with notes) if:

• Non-critical console warnings
• Minor styling issues
• Slow load times

Visual Inspection Guidelines

When viewing the screenshot, check for:

1. Layout integrity - Is content properly positioned?
2. Text readability - Is text visible, correct font/size?
3. Interactive elements - Are buttons/links visible?
4. Error states - Any error messages displayed?
5. Responsive fit - Does content fit the viewport?

Example Usage

When invoked as /visual-qa:

Running Visual QA checkpoint...

Checking console for errors...
Found 0 errors, 2 warnings.

Taking screenshot...
Screenshot saved to /tmp/tabz-screenshot-123.png
[Viewing screenshot...]

Page appears to load correctly. Navigation visible, content renders.

Checking page info...
Title: "My App - Dashboard"
URL: http://localhost:3000/dashboard
Status: Complete

Result:
{
  "passed": true,
  "issues": [
    {"severity": "warning", "message": "React DevTools warning", "type": "console"}
  ],
  "screenshots": ["/tmp/tabz-screenshot-123.png"],
  "console_errors": 0,
  "summary": "Page loads correctly. Minor console warnings only."
}

Checkpoint result written to .checkpoints/visual-qa.json

Troubleshooting

No MCP connection:

mcp-cli tools tabz  # Should list tabz_* tools

Screenshot fails:

• Ensure Chrome tab is focused
• Check TabzChrome extension is active
• Verify localhost:8129 backend is running