General: Chrome Browser Validation
Chrome browser automation via mcp__claude-in-chrome__* tools for visual validation.
When to Use
| Condition | Action |
|---|---|
/p spec doc has Chrome Validation Tab URL | Use that URL after completing tasks |
| User provides URL in prompt | Use that URL after completing task |
| User says "check chrome" but no URL | Ask user for the URL |
| No URL mentioned | Do NOT offer Chrome validation |
Mandatory Workflow
1. Get Tab Context (Always First)
code
mcp__claude-in-chrome__tabs_context_mcp
2. Get URL
Priority: 1) /p spec doc URL, 2) User prompt URL, 3) Ask user
3. Navigate and Validate
code
mcp__claude-in-chrome__navigate (url, tabId) mcp__claude-in-chrome__computer (action: "screenshot", tabId)
4. Report Findings
- •What you observed
- •Whether implementation matches expectations
- •Any issues found
Available Tools
| Tool | Purpose |
|---|---|
tabs_context_mcp | Get available tabs (call first) |
tabs_create_mcp | Create new tab |
navigate | Navigate to URL |
computer | Screenshot, click, type, scroll |
read_page | Get accessibility tree |
find | Find elements by description |
form_input | Set form field values |
javascript_tool | Execute JS in page |
read_console_messages | Read browser console |
get_page_text | Extract text content |
Base URL
App runs at: http://localhost:5173/ - assume dev server is always running.
Common Mistakes
- •❌ Validate without URL trigger → ✅ Only validate when URL provided
- •❌ Skip
tabs_context_mcp→ ✅ Always call it first - •❌ Guess URLs → ✅ Use exactly what user/spec provides
- •❌ Attempt login → ✅ Assume tabs are authenticated
- •❌ Reuse old tab IDs → ✅ Tab IDs change between sessions
When NOT to Use
- •User didn't provide URL and didn't mention Chrome
- •Backend-only changes
- •Changes verifiable via type checking alone