[andrewcai8]

Merges 481 upstream commits, through upstream's newest commit with green CI. 40 files conflicted.

Conflict resolutions keep upstream's changes and re-apply the fork's behavior on upstream's new code, much of which moved to Effect.fn:

• Scheduled tasks (ScheduledTaskService): the target column, cloud fires and their crash resume, and a cloud-only host defaulting to a cloud machine and refusing local. Upstream's runNow returns threadId; a cloud fire reports null, because its chat lives on its machine.
• Server config (ws.ts, cli/config.ts): no Scratch on a cloud-only host, environment-control and Home fields in the server config, and the client home patch stripped.
• Providers:
  • The Claude driver keeps the host probe (setup-token usage, claude auth status fallback) inside upstream's capabilities cache.
  • The Codex driver keeps the fork's host login and environment.
  • The Codex probe keeps the fork's 30s budget for app-server startup on EFS, minus upstream's discovery margin.
  • Registry hydration keeps layerWithDrivers with upstream's subscribe-first settings watcher.
  • The Claude adapter drains wakes when the process ends, alongside upstream's nativeWakeStarted.
• Orchestration and MCP: Home routing and the fork's MCP helpers, the failed-turn child rule in RunExecutionService, the cloud-only launch refusal, imported native sessions, and the fork's skipping of task-notification records, now in upstream's isHiddenClaudeRecord.
• Clients:
  • The cloud Run-on picker sits beside upstream's "Add to another machine".
  • Cloud boxes connect only while wanted, including when a compatibility error clears.
  • The mobile usage-limit card has "Switch account" on upstream's Button.
  • Held cloud sends, settlement scoping and provisioned skills are kept.

Follow-ups the merge needed:

• Upstream's new t3code(prefer-effect-fn) lint rule (feat(lint): require Effect functions for reusable generator wrappers pingdotgg/t3code#22000) flagged 64 functions in fork-only code. They are now Effect.fn, or Effect.fnUntraced where the function already ends in its own withSpan.
• Upstream's CodexUsageHistory service is provided to two fork usage tests, and to the usage upload script as an empty in-memory history.
• ws.test.ts mocks the fork's HomeService and EnvironmentControl. An MCP test's registry mock gains refresh, which the fork's auto-pick calls.
• The Codex probe-budget tests advance by the fork's 30s budget. ForkMigrations.test.ts lists upstream migrations 65 and 66.

Verification:

• Typecheck is clean for contracts, shared, client-runtime, provider-core, server, web, mobile, desktop and scripts.
• Repo lint has no errors.
• Server tests for every file both sides changed, plus the fork's cloud, account-switch, migration and usage tests, pass (1888 tests). The rewritten modules' tests pass too (about 2600).
• Web unit tests (7369) and mobile unit tests (3537) pass.
• Locally, a few tests fail only on this Mac:
  • two macOS native race tests and a /var symlink path test, all identical to upstream;
  • Unix socket path length under a long TMPDIR;
  • a Codex install test on a case-insensitive filesystem;
  • one Claude adapter steer test that times out on this Mac but already passes in the fork's CI.

🤖 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	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	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	21.2 KiB	21.2 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 5f695bb · PR result: 1f10496 · 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.

[andrewcai8]

Release Smoke failed with ERR_PNPM_UNUSED_PATCH for expo-modules-core@58.0.22. Version 58.0.23 passed the minimum release age today, and upstream's CI fails the same way. I applied upstream's fix from pingdotgg#22343, which pins expo-modules-core: 58.0.22 in overrides. Release smoke passes locally. Mobile Native Static Analysis was cancelled by the same run, not failed.