| name | e2e-page-tests |
| description | When layout, functionality, or conditions change in apps/web or apps/management-web, add or update the corresponding E2E (Playwright) test so page behavior stays covered. |
| version | 1.3.4 |
E2E Page Tests (Web and Management-Web)
Testing requirement policy lives in feature-implementation-testing. This skill focuses on how to add or update E2E coverage for web and management-web changes.
Current E2E bar: Confident. Use this skill when you change layout, functionality, or conditions (e.g. redirects, auth checks, visibility, error states) in apps/web or apps/management-web. Always add or update an E2E test so the change is covered and regressions are caught. Updating E2E tests is required for functional changes to web or management-web; it must not be treated or listed as optional in plans or implementation reports.
When to add or update a test
- Layout changes – New or moved sections, nav, headings, or structure on a page.
- Functionality changes – New or changed forms, buttons, links, or user flows (e.g. create bucket, edit message, login).
- Condition changes – New or changed redirects, auth guards, visibility rules, or error/empty states.
For CRUD and permission-gated flows, also apply e2e-crud-state-matrix and, for permission-gated pages, e2e-permission-actor-matrix.
For query-param state behavior, also apply e2e-url-state-contracts.
For test-title/step readability and report behavior, also apply e2e-readability.
If the change is in web, add or update a spec in apps/web/e2e/. If it is in management-web, add or update a spec in apps/management-web/e2e/. When adding a new spec file, update the corresponding E2E spec order file so the full report stays in conceptual order (see e2e-report-order skill).
End your response with exact E2E verification commands. Whenever you add, change, or remove E2E specs or change UI that E2E tests cover, you MUST end the response with a fenced bash block containing the EXACT make command(s) the user needs to run to verify the change (e.g. make e2e_test_web_report_spec SPEC=e2e/settings-bucket-owner.spec.ts). See response-ending-make-verify skill. No exception.
Deterministic outcome policy (required)
- Each E2E test must have one deterministic expected outcome for the mode/config it runs under.
- No dual-condition or mode-agnostic tests. Do not write a single test that accepts "either outcome A or outcome B" (e.g. "either success message or error message") depending on config or env. Such tests hide regressions and violate single-outcome policy.
- Do not permit dual outcomes such as "either error message A or redirect B." Use one assertion path per test.
- Do not use
Promise.race(...) or broad regex patterns to accept multiple incompatible outcomes.
- Separate spec files per mode (avoid
test.skip()). When behavior is mode-dependent (e.g. AUTH_MODE=admin_only_username vs AUTH_MODE=admin_only_email or user_signup_email), use separate spec files per mode (e.g. settings-bucket-owner.spec.ts for default, settings-bucket-owner-admin-only-email.spec.ts for admin-only-email) and run each spec only under the config where that outcome is guaranteed. Do not use test.skip() to gate mode-dependent tests so that "skipped" is not necessary; each file has a single outcome per test and runs in one config.
- A test should fail when the non-target mode behavior appears.
Timeout increases are almost never the fix (required)
- Do not resolve failing or flaky E2E tests by increasing timeout values (
test.setTimeout(...), test.slow(), Playwright timeout, expect.timeout, or per-call { timeout: ... }) as the first response.
- In nearly all cases, timeout failures indicate a readiness, determinism, or assertion issue, not that tests should simply wait longer.
- Fix root cause first:
- add explicit destination-load verification after navigation (URL and destination-specific visible element),
- assert the specific success/error state tied to the user action,
- use deterministic setup data (especially for one-time tokens and links),
- remove arbitrary sleeps in favor of condition-based waits.
- If a timeout increase is truly unavoidable, treat it as a rare exception: document why inline in the spec and keep the increase minimal.
Where tests live
| App | Specs directory | Config |
|---|
| apps/web | apps/web/e2e/ | apps/web/playwright.config.ts (default) |
| apps/web (signup) | apps/web/e2e/ | apps/web/playwright.signup-enabled.config.ts |
| apps/management-web | apps/management-web/e2e/ | apps/management-web/playwright.config.ts |
- Default web E2E uses
AUTH_MODE=admin_only (signup disabled). For signup-enabled auth flows (signup, forgot-password, reset-password), use make e2e_test_web_signup_enabled; it starts Mailpit via infra/docker/e2e/docker-compose.yml and uses the signup-enabled config. make test_clean removes Mailpit with other test containers. See docs/testing/E2E-PAGE-TESTING.md (“E2E and AUTH_MODE”).
- Use the deterministic E2E seed for data (e.g.
e2e-bucket-owner@example.com / Test!1Aa for web bucket-owner; management-web login is by username e2e-superadmin and password Test!1Aa). See docs/testing/E2E-PAGE-TESTING.md.
- API gate: E2E Make targets run API integration tests first; if they fail, Playwright does not run.
- Current startup model: Playwright
webServer now auto-starts the required API + web apps on dedicated E2E ports in production-like mode (build + start), so manual app startup is not part of normal E2E runs.
Rate limiting and auth in E2E
- Rate limiting disabled in test: In test, rate limits are effectively disabled (very high limit) so E2E and API integration tests never hit 429. Do not add tests (E2E or API) that assert rate-limit behavior (e.g. 429 after N requests); see api-testing skill.
Placeholder plans
Page-level coverage is tracked in .llm/plans/active/e2e-page-tests/ (e.g. web-03-dashboard.md, mgmt-06-bucket-detail.md). When you add or expand a test for a page, consider updating the corresponding placeholder with the new scenarios or marking it as implemented.
Quick reference
- Default execution strategy (agent/sandbox): use the leanest E2E command that validates the change first (single spec/report-spec target), then broaden only when required.
- Most targeted web E2E run:
make e2e_test_web_report_spec SPEC=e2e/<spec>.spec.ts
- Most targeted management-web E2E run:
make e2e_test_management_web_report_spec SPEC=e2e/<spec>.spec.ts
- Run E2E (web only):
make e2e_test_web
- Run E2E (management-web only):
make e2e_test_management_web.
- Run E2E (both):
make e2e_test.
- Run report-focused home smoke (auto-opens HTML reports, captures step screenshots):
make e2e_test_home_report.
- Signup-enabled web E2E (signup + forgot/reset auth pages; starts Mailpit, report to
web-signup-enabled/): make e2e_test_web_signup_enabled.
- Admin-only-email web E2E (scoped report):
make e2e_test_web_admin_only_email_report_spec SPEC=e2e/<spec>.spec.ts (e.g. SPEC=e2e/settings-bucket-owner-admin-only-email.spec.ts). Full suite: make e2e_test_web_admin_only_email.
- Docs: docs/testing/E2E-PAGE-TESTING.md.
Report mode uses a custom reporter (scripts/e2e-html-steps-reporter.ts) so each
step screenshot is shown with its full "Step description" in an expandable block
directly below the image; when the capture helper attaches a "Step URL", the report
shows that URL in the same block. The report UI includes a fixed top-right indicator
showing which test you are viewing (e.g. 3 / 12) and fixed bottom-right nav (prev/next
test, shot, error).
Report title and meta title (required)
Every E2E HTML report must use a descriptive title in the page heading and <title> (browser tab), not a generic label like "E2E step report". The required pattern is:
- Web (default or scoped):
E2E Web Report - e2e/<spec>.spec.ts (or comma-separated spec paths when multiple).
- Web admin-only-email: same pattern; output dir must be
web-admin-only-email so the reporter uses the Web Report title and appends E2E_REPORT_SPEC.
- Web signup-enabled:
E2E Web Report – AUTH_MODE=user_signup_email, MAILER_ENABLED=true with optional - e2e/<spec>... when scoped.
- Management-web:
E2E Management Web Report - e2e/<spec>.spec.ts.
The reporter derives the title from PLAYWRIGHT_HTML_OUTPUT_DIR (must end with one of: web, web-admin-only-email, web-signup-enabled, management-web) and E2E_REPORT_SPEC (spec path or comma-separated list). When adding or changing Make report targets, always set E2E_REPORT_SPEC and use one of these output dirs so the generated report has the correct heading and meta title.
Screenshot naming policy (QA-readable)
When adding or updating screenshot steps in E2E specs:
- Use
actionAndCapture / capturePageLoad from e2e/helpers/stepScreenshots.
- Keep screenshot capture report-focused (
E2E_STEP_SCREENSHOTS=true runs such as e2e_test_home_report).
- Use very descriptive step labels that explain expected visible UI outcome, not short action codes.
- Prefer long, explicit labels over short ambiguous labels so QA can infer expected state from filename alone.
Good label examples:
User navigates to the home route and is redirected to the login page for an unauthenticated session.The dashboard screen is visible with the primary heading after successful login.
Avoid labels like:
