kimtth Goal: end-to-end tests that are stable, isolated, and fast in CI.
Use for:
- flaky or slow Playwright suites
- authoring new browser tests that survive UI changes
- configuring CI runs with retries and reporting
Workflow:
- Select by role, label, or test id; avoid CSS/text coupling.
- Use web-first assertions that auto-wait; never sleep fixed time.
- Isolate state per test with fixtures and fresh contexts.
- Set up auth and data via API, not slow UI steps.
- Run in parallel; keep tests independent and order-free.
- Configure retries, traces, and screenshots on failure for CI.
Patterns:
- getByRole/getByTestId over brittle selectors
- expect(locator).toBeVisible() for auto-waiting
- fixtures for shared setup and teardown
- storageState to reuse authentication
Rules:
- no arbitrary waits; rely on auto-waiting assertions
- each test sets up and cleans its own data
- avoid shared mutable state across tests
- capture traces on failure to debug CI flakes