Prompt file imported from tredgate/tredgate_training_web_app (.github/prompts/t17-e2e-smoke.prompt.md). Copyright stays with the author.

Task 17 — E2E Smoke Validation (On-Demand)

Layer: Validation (not implementation) · Depends on: Any subset of T9–T14 marked ✅ in STATUS.md · Parallelizable: No — runs alone against a live dev server Trigger: On-demand only. Do NOT run automatically when a task is marked ✅. The user will explicitly request this task.

Objective

Validate the running app against its advertised behaviour using the Playwright MCP server. This is a smoke pass, not a test suite. Confirm the golden paths work for each role, confirm role-gated UI matches architecture §12, confirm data-testid registry is wired to the DOM. Catch integration regressions unit tests can't see.

This task does NOT:

• Replace the student-facing Playwright training tests (those are written by students).
• Modify any source code. If bugs are found, a follow-up fix prompt is created — Vera reviews it, the user approves it, another agent implements it.
• Do pixel-perfect visual checks, content assertions beyond testIds, or performance benchmarking.

Prerequisites (check before starting)

1. .github/prompts/STATUS.md shows at least one of T9–T14 as ✅ done. If all Phase 3 tasks are still pending, abort and report back — there's nothing to smoke-test yet.
2. npm install has run (node_modules present).
3. Dev server starts cleanly via npm run dev — capture the local URL (usually http://localhost:5173).
4. src/shared/testIds.ts exists and is importable — the agent reads testIds from this registry, never hardcodes them.

If any prerequisite fails, stop and report which one.

Role Matrix — Login + Sidebar Visibility

For each of the three seeded users, log in and assert sidebar contents match architecture §11.

(Exact credentials — confirm against src/data/seed.ts. If seed uses different values, use those.)

For each role, after login:

• Assert URL is /dashboard.
• Assert data-testid="dashboard-page" is present in DOM.
• Assert each expected sidebar item has its data-testid from the registry.
• Assert disallowed sidebar items are absent from DOM (not just hidden — per §12 "hide, don't disable").

Golden Paths — One per Role

Run only the paths for which the relevant page task is ✅ in STATUS.md. Skip gracefully with a "not ready" note if the page isn't implemented yet.

Tester golden path — Report a Defect

Requires T11 ✅.

1. Login as tester.
2. Click sidebar "Defects" → assert data-testid="defect-list-page".
3. Click "Report Defect" button.
4. Walk through all 4 wizard steps, filling minimum required fields. Use any valid project and environment.
5. Submit on Step 4 → assert success toast appears (data-testid="toast-success").
6. Assert navigation to /defects/{newId} — detail page loads with the title just entered.
7. Back to defect list → assert the new defect row is present (search by title).

QA Lead golden path — Assign and Start a Defect

Requires T11 ✅.

1. Login as qa_lead.
2. Navigate to /defects → open any defect with status new.
3. Assert "Assign" transition button is visible (data-testid="defect-detail-btn-assign").
4. Click "Assign" → modal opens (data-testid="modal-assign-defect").
5. Select an assignee → confirm.
6. Assert status badge updated to assigned, toast shown, history timeline has a new entry.

Admin golden path — Settings + User Role Edit

Requires T13 ✅ and T14 ✅.

1. Login as admin.
2. Click sidebar "Settings" → assert data-testid="settings-page".
3. Assert Data Management card renders (reset / clear / export / import buttons all have correct testIds — do NOT click destructive buttons).
4. Navigate to /team → assert data-testid="team-list-page".
5. Click a non-admin user row → assert data-testid="user-detail-page".
6. Assert "Edit Role" button visible (data-testid="user-detail-btn-edit").
7. Click it → assert role Select appears. Do NOT save (no state mutation needed for smoke).

data-testid Coverage Spot-Check

For each page visited during the run, assert the page-level testId from src/shared/testIds.ts is present in the DOM. If a testId is declared in the registry but missing from the rendered page, that's a bug — log it.

List of page-level testIds to check (whichever were visited): dashboard-page, project-list-page, project-detail-page, project-form-page, defect-list-page, defect-detail-page, defect-form-page, testplan-list-page, testplan-detail-page, testplan-form-page, testrun-execution-page, team-list-page, user-detail-page, reports-page, settings-page, profile-page.

Output Format

Structured report — no prose, no filler. Return exactly this shape:

## E2E Smoke Report — {timestamp}

Dev server: {url}
Tasks covered: T9, T11, T13, T14   ← whichever were ✅ at run time

### Role matrix
- [x] tester   — login ok, sidebar ok
- [x] qa_lead  — login ok, sidebar ok
- [ ] admin    — FAIL: sidebar missing "Settings" item (expected testid sidebar-item-settings)

### Golden paths
- [x] Tester: Report a Defect         — ok (defect #42 created)
- [ ] QA Lead: Assign and Start       — FAIL at step 4: modal did not open (testid modal-assign-defect not found in DOM). Screenshot: {path}
- [~] Admin: Settings + Role Edit     — skipped (T13 not yet ✅)

### testId coverage
- [x] All visited page-level testIds present in DOM.
- [ ] Missing: defect-detail-card-assignment (declared in registry, absent on /defects/3)

### Summary
2 failures, 1 skipped, 5 passed.

Use [x] pass, [ ] fail, [~] skipped.

Bug-Found Protocol

For each failure:

1. Capture a screenshot (Playwright MCP provides this) — save to .playwright-mcp/smoke-{timestamp}/.
2. Do NOT patch the bug yourself. Instead:
3. Create a new prompt file: .github/prompts/tN-fix-{short-name}.prompt.md (where N is the next free task number). Include:
   • What broke (exact testId / selector, expected vs actual).
   • Which task introduced the area (e.g., "affects T11 DefectDetail").
   • Reproduction steps.
   • Screenshot path.
4. Add a row to .github/prompts/STATUS.md for the new fix task with status ⬜ pending, Notes: "found by T17 smoke on {date}".
5. Tag Vera in the report: "→ fix prompt created: tN-fix-{name}. Review before implementing."

Do NOT

• Do not run this task automatically after another task is marked ✅ — wait for explicit user request.
• Do not modify any file under src/.
• Do not write or modify Playwright test files under tests/ or e2e/ — those are for students.
• Do not click destructive buttons (Settings → Reset / Clear / Delete modals).
• Do not make content assertions beyond testIds and toast visibility. (No "assert heading text equals X".)
• Do not take pixel-diff snapshots.
• Do not run performance or accessibility audits (separate concern).
• Do not mark this task ✅ in STATUS.md on completion — it's a recurring on-demand check, not a one-shot task. Leave it as ⬜ pending and append the run timestamp in Notes.

STATUS.md entry

Add this row once (if not already present):

| T17  | E2E Smoke Validation (on-demand)   | ⬜ pending  | Any of T9–T14 ✅| —          | —          | —       | Runs on demand; never auto-triggered |

After each run, update the Notes column: Last run: {date} — {n} pass / {n} fail / {n} skip.