Browser QA

You're about to drive a real browser to verify a feature. Use Playwright MCP. It's the only browser stack that pays off in practice — fast, scriptable, snapshot-aware, and failures are at the prompt layer so you can recover with smarter retries instead of restarting the runtime.

First-time setup — DO THIS BEFORE ANYTHING ELSE

You will be blocked mid-task if you wait until you need a permission to ask for it. Surface anything missing in the very first message of the QA phase:

1. Verify Playwright MCP works by calling mcp__playwright__browser_close once. It's a no-op if there's no open tab, and surfaces "browser already in use" lock errors early so you can pkill -f "mcp-chrome-" before they bite you mid-task.
2. Confirm your dev server is up. A quick curl -s -o /dev/null -w "%{http_code}\n" http://localhost:<port>/ saves a "Page Title: ECONNREFUSED" round-trip.

If anything is missing, surface it immediately — don't start QA half-blind.

Why Playwright? It's fast, scriptable, and snapshot-aware, and its failures surface at the prompt layer so you can recover with smarter retries instead of restarting the runtime. Headless screenshots are dense and clinical, which is exactly what a reviewer wants. Stick to one stack and learn its sharp edges.

Playwright MCP failure modes — all fixable from your side