test(agent): pin the admin dashboard page contract - AdminDashboardPage

[dmitrii-f-t27]

Closes #1594

First suite for entrypoints/app/admin-dashboard/AdminDashboardPage.tsx (199 lines, unmodified). The history hook answers its initial empty map (it populates through effects a static render never runs), so the page's static answer is the EMPTY dashboard: three tests pin its header and approvals slots (stubbed with named markers - their own suites own their contracts), the Audit Trail section with its empty-state card (the shield icon, the note, and the run-count line pinned as ABSENT by its own shape - the delete dialog's fixed sentence also carries the words 'recorded run', measured), and the delete AlertDialog in its controlled-closed shape (title, description with the empty task name, Cancel and Delete). The storage seam is the repository's in-memory item shim. Not pinned: the populated audit trail (it rides the effect-populated hook), the delete flow and the toasts.

Verified: suite alone 3/3; the admin-dashboard directory one process green; broken copy (the empty-state title reworded, the section title reworded - valid syntax) 2 catcher tests fail, restored byte-identical; biome clean after one auto-format; tsgo with 0 mentions of this file.

Spec: the issue is the executable contract for a test-only change; the suite is the pin.

Dmitrii Fedorov (@dmitrii-f-t27)

[tri-claw]

Summary by t27-bees

This PR adds a contract test for AdminDashboardPage that pins its empty-state static render. The test stubs the history hook (which populates via effects that don't run in renderToStaticMarkup), the storage layer, and child components, then asserts on header/approvals slots, the Audit Trail empty-state card, and that the delete dialog renders nothing when closed.

Worth a look

• trios/agent-server/apps/agent/entrypoints/app/admin-dashboard/AdminDashboardPage.test.ts:47–61 — The @wxt-dev/storage mock returns a hardcoded fallback only for local:tool-approval-config; if the component starts reading other keys without fallbacks, the test will silently pass with null instead of surfacing the missing configuration.
• trios/agent-server/apps/agent/entrypoints/app/admin-dashboard/AdminDashboardPage.test.ts:62–66 — useExecutionHistoryByConversation is mocked to return {}, but the component may expect { data: [], isLoading: false } or similar shape; a shape mismatch would only appear as a runtime error in the component, not a test failure here.
• trios/agent-server/apps/agent/entrypoints/app/admin-dashboard/AdminDashboardPage.test.ts:68–83 — The AlertDialog mock renders children only when open is true. This mirrors the "group's dialog stand-ins" per the comment, but diverges from the real Radix AlertDialog (which renders portal content regardless of open and controls visibility via CSS). If the real component ever relies on the dialog being in the DOM while closed, this test won't catch it.
• trios/agent-server/apps/agent/entrypoints/app/admin-dashboard/AdminDashboardPage.test.ts:108 — expect(m).not.toContain('recorded runs across') pins the absence of the run-count line, but the comment notes the delete dialog's description also contains "recorded run". If the dialog ever renders while closed (e.g., a bug in the open prop), this assertion could flakily pass/fail for the wrong reason.
• trios/agent-server/apps/agent/entrypoints/app/admin-dashboard/AdminDashboardPage.test.ts:95–97 — The test imports AdminDashboardPage after all mocks are registered. This works because mock.module hoists, but it couples test execution order to the mock setup; if another test file in the same process imports the component first (e.g., via a shared setup), the mocks may not apply.

---

_{Read: 1 files, +132 -0, the whole diff. Head db8044d. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.}

[github-actions]

✅ Tests passed — 3811/3881

Suite	Passed	Failed	Skipped
✅ agent	409/409	0	0
✅ build	9/9	0	0
✅ cdp-protocol	5/5	0	0
✅ eval	93/93	0	0
✅ server-agent	280/280	0	0
✅ server-api	2189/2258	0	69
✅ server-browser	6/6	0	0
✅ server-integration	10/11	0	1
✅ server-lib	279/279	0	0
✅ server-pglive	166/166	0	0
✅ server-root	68/68	0	0
✅ server-sim	8/8	0	0
✅ server-skills	31/31	0	0
✅ server-tools	244/244	0	0
✅ shared	14/14	0	0

View workflow run

[dmitrii-f-t27]

Gate check before merge (the fixed head):

Tests: 19/19 green on this exact SHA - incl. Tests / agent (the first push broke the sibling Skills and ScheduledTasks suites: an unconditional alert-dialog passthrough rendered their CLOSED alerts' content and flipped their not.toContain pins; the stand-in's root now honours the open flag exactly like the group's dialog registrations, and the mixed order over admin-dashboard + skills is green 10/10 - measured locally) and Tests / server-pglive (no rerun needed).

runner / Typecheck - GREEN, 0 errors, 0 mentions of this PR's file.

runner / Biome - red from base, 0 mentions of this PR's file, 0 new findings.

Dmitrii Fedorov (@dmitrii-f-t27)