kandev/docs/plans/utility-agent-profiles/task-05-utility-profile-e2e.md

id, title, status, wave, depends_on, plan, spec

id	title	status	wave	depends_on	plan	spec
05-utility-profile-e2e	Prove utility profile workflows	pending	4	03-update-utility-consumers 04-settings-profile-pickers	plan.md	../../specs/agents/requirements/utility-agent-profiles.md

Task 05: Prove utility profile workflows

Intent

Prove persisted profile selection, custom-agent editing, stale-profile repair state, and equivalent phone reachability against the production E2E build.

Acceptance

• Desktop E2E saves/reloads a default profile and action override, creates/edits a custom utility agent with a profile, and asserts legacy agent/model controls are absent.
• A disabled saved profile remains visible as unavailable with repair guidance and is not offered as a runnable choice.
• Mobile E2E completes the default/action selection flow, opens custom edit, verifies touch reachability/viewport containment, and asserts no horizontal document overflow.

Files likely touched

• apps/web/e2e/tests/settings/utility-agents.spec.ts
• apps/web/e2e/tests/settings/mobile-utility-agents.spec.ts
• apps/web/e2e/helpers/api-client.ts
• apps/web/e2e/pages/settings-page.ts

Dependencies

Tasks 03 and 04.

Parallelism

Parallel-safe with task 06 after product behavior is complete: this task owns E2E files while task 06 owns documentation. Sequential execution remains the default.

Inputs

• Spec: custom/default/override, stale profile, reload, and phone Scenarios.
• Plan: E2E Tests and Mobile design contract.
• Existing patterns: current desktop and mobile-chrome utility-agent specs, agent-profile API seeding helpers, and managed production-build E2E runner.

Verification

cd apps/web && pnpm e2e:run tests/settings/utility-agents.spec.ts
cd apps/web && pnpm e2e:run --project mobile-chrome tests/settings/mobile-utility-agents.spec.ts

Confirm each command discovers the intended test count. The managed runner rebuilds backend/web artifacts and tears down isolated runtime data.

Output contract

Report scenarios, discovered/passed test counts, rendered mobile containment evidence, cleanup, files changed, blockers, risks, and synchronized task/plan status. Do not treat API-only assertions as E2E proof.

Results

The focused unit/component suites pass. The managed Playwright runner was started twice but did not emit a final discovery/result summary in this workspace, so production E2E proof remains pending for follow-up.