testing-with-playwrightlisted

# Testing with Playwright Extend the project's actual Playwright setup rather than inventing a parallel one. See [enforcing-safety-baseline](../enforcing-safety-baseline/SKILL.md) for the evidence rule this skill leans on hardest: a test result is only real once it has actually run. ## Workflow 1. **Discover existing setup.** Read `playwright.config.*`, existing `tests/`/`e2e/` structure, fixtures, page objects, and helper utilities already in use. Match the existing patterns instead of introducing a new style. 2. **Reuse fixtures and helpers** for auth, test data setup/teardown, and navigation instead of duplicating that logic per test. 3. **Prefer resilient selectors**: role/label/text-based locators (`getByRole`, `getByLabel`, `getByText`) over brittle CSS/XPath tied to implementation details, unless the project's existing convention already uses `data-testid` — then follow that convention. 4. **Never use arbitrary sleeps.** Use Playwright's built-in waiting (auto-waiting assertions, `waitForResponse`, `waitForURL`) instead of `waitForTimeout` as a synchronization mechanism. 5. **Cover the happy path** for the requested flow, then the relevant error/edge cases (invalid input, permission denied, empty states) — not exhaustive combinatorics, just what's realistically load-bearing. 6. **Check multi-tenant isolation when applicable**: if the flow touches tenant-scoped data, add a negative test proving user/tenant A cannot see or act on tenant B's data — don't assume isolati