playwright-core

Use role-based locators like getByRole instead of CSS or XPath selectors, rely on web-first assertions such as expect(locator).toBeVisible() that auto-retry, and never use page.waitForTimeout(). Isolate every test with no shared state and set retries to 2 in CI and 0 locally.

Use page.route() to intercept network requests and fulfill them with custom responses. Mock only external third-party services like payment gateways or email providers, never your own application's API, so tests exercise real application behavior.

Yes, the @axe-core/playwright package integrates axe-core scans into Playwright tests via AxeBuilder. You can run full-page scans, scope scans to specific regions, filter by WCAG tags like wcag2aa, and disable rules for tracked known issues.

Yes, the built-in request fixture provides an APIRequestContext that makes HTTP calls without launching a browser. It supports REST and GraphQL endpoints, custom headers, authentication tokens, and can be configured with a dedicated project in playwright.config.

Flakiness usually comes from hardcoded waits, non-retrying assertions on resolved values, shared state between tests, or race conditions with animations. Replace waitForTimeout with web-first assertions, isolate test data, and use traces set to on-first-retry to diagnose failures.

Perform login once in a setup project and save the session with storageState, then reuse that state across tests to skip repeated logins. For API-seeded scenarios, create users via the request fixture and clean them up after each test.