feat(mobile): manage scheduled tasks across environments

[juliusmarminge]

What changed

Add mobile scheduled-task management under Settings and the command palette. The list spans connected environments and uses the existing environment/project filters. Create and edit tasks in dedicated screens with the same model and branch pickers used by new threads, including model options and origin selection.

The editor supports composer dictation, bounds long prompt fields, locks inputs during saves, and confirms before discarding changes. It preserves valid fractional and long intervals and opens legacy sub-minute intervals at the one-minute minimum with an explicit warning before saving. The iOS back action checks for unsaved work before starting navigation. Task rows show readable intervals and upcoming run times, with pause, resume, run-now, and delete actions.

Why

Mobile needs a way to manage recurring work across environments without editing schedules through an agent. Reusing the existing pickers keeps model and branch selection consistent with ordinary tasks.

Based on t3code/codex-turn-mapping, which supplies the scheduled-task contracts and server operations. Edit saves use a new optional requireExisting precondition, checked atomically by the server, so a stale editor cannot recreate a task deleted by another client. Other upsert callers retain their existing behavior. This protection requires the updated server. No dependency changes.

Verification

• Mobile, server, and contracts TypeScript checks passed.
• 131 focused tests passed across schedule validation and formatting, dirty-state detection, duplicate weekday normalization, settings filters, branch selection, model options, dictation, and server-side stale-save protection.
• Changed-file formatting and lint passed, with existing hook-dependency warnings in the shared picker files.
• Reviewed the full diff and shared picker callers. React Doctor reported no errors; its warnings concern component size, safe-area padding, small array lookups, and the shared symbol export.

UI verification

The maintainer verified the UI on their device and explicitly waived screenshot evidence. Agent verification was code-only, as requested. No screenshots or recordings were captured.

Authored with GPT-6 via Codex.

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This XXL PR adds scheduled-task creation and management across mobile and web, including agent execution controls and server persistence changes, while also refactoring shared branch-picker behavior. It changes defaults and affects multiple existing production paths, so the scope and runtime impact warrant human review.

^{You can add or adjust custom eligibility rules. Learn more.}

[github-actions]

Thread transfer impact

⚠️ The latest CI run did not produce a thread transfer result for f5a4b7f.

This comment will update automatically after the next completed run.

[macroscopeapp]

All clear

^{Posted via Macroscope — Effect Service Conventions}

[macroscopeapp]

All clear

^{Posted via Macroscope — Effect Service Conventions}

[macroscopeapp]

All clear

^{Posted via Macroscope — Effect Service Conventions}

[macroscopeapp]

Correction: the review found no Effect service convention violations in the selected files.

^{Posted via Macroscope — Effect Service Conventions}