playwright-testerlisted

# Playwright Tester Write tests once, run them forever for free. ## The economics An AI driving a browser live pays tokens for every click and keeps nothing; screenshot-based checking is worse, because images are token-heavy and the result is a vibe rather than a gate. This skill is the third option: the AI authors tests, Playwright runs them. Tokens are spent once on *what to assert*, never on pressing the buttons, and the output survives the conversation, runs in CI unchanged, and grows into regression coverage. **When something else fits better:** a one-shot "just look at this page and tell me what you see" is a live-browser job - use the browser tools. If the user asks for Playwright's own agents (`npx playwright init-agents`, the Playwright MCP server, `@playwright/cli`), point them there rather than duplicating it. ## Setup Check for `playwright.config.ts` (or `.js`) and a `@playwright/test` devDependency first - most projects have one, and re-scaffolding over it risks clobbering config. If nothing is there: ```bash npm init playwright@latest # scaffolds config + tests/ + optionally a CI workflow # or, non-interactively: npm install -D @playwright/test && npx playwright install chromium ``` `assets/playwright.config.ts` is a working starting config. Copy it and adjust the dev-server command and port rather than writing one from memory. **If more than one agent may test this repo at once, set `PW_PORT` to a per-agent value before running anything.** The d