[jayleaton]

Stacked on #224, which is stacked on #222.

Scheduled tasks had two web UIs. Settings → Scheduled tasks listed every task, but editing an agent task hid the agent and showed its cron as read-only ("Edit it from the Agents board"). Agents → Scheduled had a second list and editor that only knew about agent tasks. Mobile had no way to pick an agent or edit a cron. At phone width, #222's Active/Archived toggle also overlapped the heading.

This PR makes one list and one editor, per the design's UI unification section:

• One list. Settings and Agents → Scheduled now render the same list. Agents → Scheduled opens with the Agents filter selected. ScheduledTaskEditor.tsx and ScheduledTasksPanel.tsx are deleted.
• Filter. All / Agents / Threads / New thread / Archived replaces the Active/Archived toggle. The filter is scheduledTaskMatchesFilter in client-runtime/scheduled-task-archive, built on the new contracts helper scheduledTaskTarget, and web and mobile share it. Archived keeps its reason, its "Deletes on" date and Restore. At 390px the filter wraps onto its own row.
• Editor. Destination gains An agent's thread with an agent picker. Saving routes through the agent's model and modes, as the old agent editor did. An existing task keeps its model and thread while its agent is unchanged, and moving a task off an agent clears the agent. The read-only "Current" cron mode is replaced by a real Cron mode with presets, time, day and time zone. Cron schedules are now described in plain words in the list.
• Entry points.
  • The agent tile's Schedule opens the shared editor with that agent filled in.
  • The thread details Automations section gets a + that opens the editor bound to that thread. As before, the section only shows once the thread has a task.
  • The Agents route's mutation controls follow permissionAtom, because it now uses the Settings list.
  • Webhook tasks no longer offer Run now there.
• Mobile gets the same filter (as a menu chip, since five segments don't fit a phone), an Agent destination with an agent picker, and an editable Cron mode.
• The cron helpers move to client-runtime/scheduled-task-cron so both clients can use them, and isoToLocalInput and localInputToIso are removed as unused.

Captures (web dev client, real seeded data)

Before	After
Settings list, light	Unified list with filter and agent badges
Archived toggle	Archived filter with Restore
Agents → Scheduled, its own list, dark	Same list, Agents preselected, dark
Settings editor on an agent task: no agent	Agent picker
Cron read-only ("Current")	Editable cron
390px: toggle overlaps heading	390px: filter wraps
Old agent editor	Thread Automations + → bound editor

The red "Doug can't run here" line in the agent-picker capture is real: that agent's profile doesn't list this dev environment. The old editor showed the same thing.

Mobile has no captures yet. Its simulator pass is still to come. It typechecks and its tests pass.

Verification

• vp test run on the touched files gives 87 passed across 6 files. These include ScheduledTasksSettings.browser.test.tsx and ThreadAutomationsPanel.permissions.browser.test.tsx, plus the logic tests in client-runtime, web settings and mobile.
• vp run typecheck in apps/mobile reports 0 errors.
• vp run check:changed: all pass (format, lint, knip ×2, typecheck of 41 packages, 30 test files, React compiler).

Implemented by Claude Opus 5.5 in Claude Code (T3 Code).

🤖 Generated with Claude Code

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	21.5 KiB	21.5 KiB	0 B (0.0%)	24.4 KiB	✅
Codex	Live turn messages	1	1	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.9 KiB	21.9 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	1	1	0 (0.0%)	8	✅

Baseline: aa65f1b · PR result: 5c42bc7 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[jayleaton]

VERDICT #226: CHANGES

Confirmed the inline finding with a deterministic service test. Reviewed head 35158bb.

Repository checks against #224 passed apart from the missing local Chromium executable; the 2 browser files subsequently passed (11 tests). The other 6 test files passed (87 tests). Format, lint, knip, typecheck and React compiler checks passed. Real web create/list verified at 390px; native mobile could not be exercised because no simulator/emulator host is available. The budget failure is inherited from #224; the fingerprint check now passes.

[jayleaton]

Fixed in 6d40a19: rebased on #224, and leaving an agent with no thread chosen now clears the agent's old thread.

[jayleaton]

VERDICT #226: MERGE

Reviewed 6d40a19 against e7d0f17. check:changed passes: formatting, lint, both knip passes, GraphQL, 41-package typecheck, 14 focused test files (196 tests), and React compiler checks.

The server clears the old binding only when profileId and threadId are explicitly null while leaving an existing agent. Omitted fields retain legacy behavior, and the same path handles Continue and Fresh session. The extended profile-service regression fails before the fix (the old agent thread is retained) and passes afterward, including a fresh non-agent launch instead of sending to the old conversation.

This layer passes review independently. The stack remains blocked by the new cleanup-reservation finding on #222.