chore: sync T3 Code upstream through 6ea01f8d

[filipgutica]

Summary

Merge 23 upstream commits through 6ea01f8d24419019588f2be589681929d8e26b0c into product main da6a067bc56db69cb5d988f5abf903df01581614, preserving upstream ancestry. PR #91's first-use Ticket repository review and staged Save/Cancel remain intact.

Changes

• Import agent-session restart, Claude subagent model attribution, desktop browser download handling, name-only project creation, and clone favicon fixes.
• Import Expo 58 / React Native 0.88.0-rc.3, scene-aware native controls, notification changes, and Android subscription widgets.
• Resolve .github/workflows/ci.yml with upstream's parallel jobs, six weighted server shards, and aggregate Check. Retain the fork's runners, timeouts, serialized package tests, and release guard. Guard APT source edits by file existence.
• Regenerate pnpm-lock.yaml from merged manifests with pnpm 11.10.0 and Node 24.13.1; frozen installation passes.
• Preserve the fork distribution/updater policy and all Workbench integrations. No Workbench business-rule conflict repair or unrelated isolation cleanup was needed.

Isolation assessment

Compared pre-PR90 c108a3077e, post-PR90 c91a29326a, and current product da6a067bc against the same pinned upstream SHA, before synchronization:

Product ref	Overlap paths	Conflict paths
Before PR #90	12	2
After PR #90	12	2
After PR #91	12	2

All share merge base 67b175a4; conflicts remain CI and the lockfile. These results do not establish reduced overlap or conflict frequency.

Isolation commits 3d006d0873, 9ef1b328e0, a4e0ba14f4, 79e1444d98, and 97ea2a041b moved RPC service acquisition, sidebar/context slots, desktop policy, and Ticket summary ownership into fork modules. PR #90 also changed notifications and demos; its full diff is not an isolation metric. Earlier isolation work already exists in the selected baseline.

Every affected shared integration was reviewed, including clean merges in CommandPalette, RPC authorization/contracts, WebSocket composition, server tests/build configuration, and release policy. The RPC service adapter survived upstream project/session changes without manual repair. Its fork delta against the common merge base shrank from 25 to 13 added lines across PR #90. This is a smaller integration footprint, not a measured reduction in sync intervention. CommandPalette, auth, and schema adapters predate PR #90. Sidebar, desktop policy, and summary extraction were not stressed by this upstream range. Package boundaries held; explicit shared seams and CI/dependency conflicts remain maintenance costs.

Verification

• Package isolation, Workbench quality and complexity/duplication gates pass.
• Contracts, server, web, desktop, scripts, mobile, and browser-tooling typechecks pass.
• Focused Workbench contracts/store/adapters/UI/updater/sync tests pass. Storage cleanup passes all 62 tests with canonical TMPDIR=/private/tmp.
• Changed provider, desktop preview/download, project, asset, path, mobile widget/notification, and relay tests pass. ProviderRegistry passes all 53 tests with a portable PATH; the normal macOS PATH adds legitimate Homebrew probes to an unchanged command-list assertion.
• Desktop build and Electron runtime verification pass locally. The dedicated Linux desktop build/Electron smoke check passes in CI on this PR head.
• Current-head CI passes lint, application typechecks, builds, web/package tests, all six weighted server test shards, release smoke, and native static analysis. Workbench quality/isolation, complexity/duplication, and all three lifecycle regression shards pass.
• All nine disposable planning, Threads, and Ticket workspace browser regressions pass, including repository review, Save/Cancel, native Thread creation, retained worktree reuse, and existing Thread relinking. Provider execution uses a fixture; these checks do not prove a real provider session.
• Independent adversarial behavior/isolation review and the focused test audit found no actionable findings at da6a067bc..7f5df57dd.
• CI YAML parses, aggregate Check covers every job, manual-resolution formatting and non-patch diff checks pass. Imported patch context is retained byte-for-byte from upstream.

To reproduce the focused lifecycle checks, run from scripts: vp exec playwright test --config workbench-regression/playwright.config.ts planning.spec.ts threads.spec.ts ticket-workspace.spec.ts. The fixture owns disposable state and services.

Limits

Exact-product Expo/RN native compilation and device runtime are unverified: this host has Command Line Tools, without simctl, CocoaPods, or Android tools. Mobile type/unit checks and current-head native static analysis pass. The fingerprint check confirms changes to both iOS and Android, requiring new native clients before compatible OTAs.

Upstream's Expo migration and native release validation report successful iOS and Android Release previews, Xcode/device smoke, and 120 Android native tests. Those records cover upstream refs; they do not verify this exact product merge. Matching product native builds and device validation remain required before mobile release.

Existing local changes and the demo on ports 7580/15620 are preserved. No release is tagged, published, or deployed by this sync.

The merge automatically queued the unchanged Mobile EAS Production workflow. That run was canceled before any job started. The pre-existing workflow has no repository guard and can submit builds or publish OTAs when credentials are configured; guarding future fork-main runs is separate follow-up work.

Implemented and reviewed with GPT-6 in Codex.

[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	13.5 KiB	13.5 KiB	−14 B (−0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	−5 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−9 B (−0.1%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.2 KiB	56.2 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	9	9	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+18 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	+7 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.4 KiB	+11 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: da6a067 · PR result: 7f5df57 · 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: 113.9 KiB
• Claude decoded thread snapshot: 114.6 KiB

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

[github-actions]

Workbench demo

Open the private 20-minute demo. Sign in to Vercel, then choose Open demo. Reopening this link in the same browser resumes its running demo; Start new demo creates a fresh one. Jira starts disconnected.

Application, runtime startup, fixtures, and native smoke: PR 7f5df57dd849ff32be77d3581ced5d9e38115876. Dockerfile, build configuration, bundle export, publisher, and protected launcher: main da6a067bc56db69cb5d988f5abf903df01581614.

Expired? Return to this link and launch again. Use synthetic content with the free OpenCode model.