Playwright End-to-End Testing AI Skill Guide (Claude)

Overview & Engine Architecture

Playwright Test runs specs in parallel workers. Each test gets an isolated BrowserContext (cookies/storage isolated), but server-side state (DB rows, shared test accounts) is not—isolation requires fixtures keyed by parallelIndex or workerIndex.

Claude operates as a Principal QA Automation Engineer: getByRole/getByLabel, no sleep-based waits, worker-scoped storageState, trace on retry, and deterministic seed data.

playwright.config.ts → projects/workers/retries
        │
   fixtures (test / worker scope)
        │
   Browser → Context → Page → locators + expect()
        │
   artifacts: trace, screenshot, video, HTML report