Playwright

Official docs

Agent rules

1. Role-based locators — getByRole, getByLabel, getByTestId; avoid brittle CSS/XPath.
2. Test isolation — each test sets up its own state; no order dependency between specs.
3. Wait correctly — prefer auto-waiting assertions (toBeVisible) over fixed sleep.
4. Ship gate — agents cannot claim UI done without running Playwright or equivalent browser verify when tests exist.
5. Project config — extend existing playwright.config.ts; do not create parallel config files.