[Yarden-zamir]

Cause

Fork Sync run 40252630761 stopped on merge conflicts in 9 files. The batch has 30 upstream commits. The ones that touch fork code:

• refactor(web): terminal focus requests name the terminal they focus pingdotgg/t3code#25009: terminal focus requests name the terminal they focus. The focusRequestId counter is gone, replaced by focusTerminal, focusActiveDrawerTerminal and related helpers.
• chore(lint): a new React effect states its reason instead of raising the budget pingdotgg/t3code#25163: an effect can state its reason instead of raising the effect budget.
• fix(server): background compactions no longer flip threads to Working pingdotgg/t3code#25122: T3's silent idle compactions stay out of the thread shell. latestRunSilent is gone.
• feat(server): count how guest sharing gets used pingdotgg/t3code#24996: guest sharing records analytics.
• Several new Effect and ESLint rules (mapSomeToAsSome, flatMapIgnoredParamToAndThen, no-unsafe-optional-chaining, …).

Conflicts and resolution

• lib/terminalFocus.ts: upstream's focus request helpers, plus the fork "terminal-thread" focus owner and its two chord helpers.
• ThreadTerminalDrawer.tsx: upstream's viewport focus targets, plus the fork props (onStatusChange, onInput, optional onAddTerminalContext).
• useChatTerminalActions.ts: upstream moved the script terminal focus after the launch. The fork rule stays there: a script that runs without its terminal does not move focus.
• ChatView.tsx, GuestShareService.ts, mobile new-task-flow-provider.tsx: imports from both sides.
• McpToolAccess.ts (adopted feat(server): work a linked environment starts stays within its link pingdotgg/t3code#16695): kept the fork side.
• react-effect-budget.ts: upstream's new rule, which still merges the fork budget file.
• pnpm-lock.yaml: rebuilt with lockfile.py. The integrity changes equal upstream's.

Fork fixes

• Terminal threads and the fork right panel (refactor(web): terminal focus requests name the terminal they focus pingdotgg/t3code#25009): ForkShellTerminal, ForkTerminalThreadScreen, ForkDraftTerminalComposer and useThreadRightPanel now use the named requests at the same call sites as upstream's chat view:
  • The main terminal focuses on mount.
  • focusMain asks for the main terminal by name.
  • The drawer and panel terminals focus as before.
  • The fork test of script focus registers real focus targets.
• Web push (fix(server): background compactions no longer flip threads to Working pingdotgg/t3code#25122): shells no longer carry silent idle compactions, so the fork's silent parameter had nothing left to skip. It is removed, together with its test. The upstream projection now holds the same rule, and a failed compaction still notifies.
• New Effect rules:
  • Fork MachinePeers.ts uses Effect.asSome.
  • In code from the adopted pull requests, PeerLinks.ts (feat(server): link to another environment pingdotgg/t3code#16655) uses Effect.asSome and McpToolAccess.ts (feat(server): work a linked environment starts stays within its link pingdotgg/t3code#16695) uses Effect.andThen, as upstream changed its own code. These are one-line changes to adopted code, so a future head update of those pull requests can conflict on these lines.
• Tests:
  • GuestShareService.fork.test.ts stubs the analytics service that upstream's guest sharing now needs.
  • AddEnvironmentDialog.browser.test.tsx no longer reads through an unsafe optional chain.

Upstream merge

Merges pingdotgg/t3code main at 0b42aa9 "chore(devices): bump agent-device to 0.21.31 (pingdotgg#25065)".

Adopted upstream pull requests

All 7 entries (pingdotgg#16653, pingdotgg#16654, pingdotgg#16655, pingdotgg#16684, pingdotgg#16695, pingdotgg#16698, pingdotgg#16719) are unchanged at their listed heads. No new merges.

Upstream changes in fork areas

32 files changed on both sides. Beyond the fixes above, the server changes are lint rewrites (Effect.asSome, pipe chains) or analytics, which do not change fork behavior.

Checks

• checks.py: typecheck, migration check, adopted pull request check and lint on all fork files pass.
• fork_tests.py: 176 fork test files pass locally (117 root and 59 web).
• Server: GuestShareService.fork.test.ts, McpToolAccess.fork.test.ts and WebPush.test.ts pass (27 tests). The other apps/server fork test files are left to Fork CI, because they stop the agent session on this VPS.

Fork repair skill changes

• checks.py runs vp fmt --check on the fork files before vp lint on a fork-sync/ branch. Fork CI's vp check failed on one unformatted fork file (WebPush.ts) that lint alone did not catch. Tested in test_scripts.py. SKILL.md says so.

[coderabbitai]

Finishing Touches: 💡 1

🛠️ Fix failing CI checks: 💡

• Commit to this branch
• Create a new PR

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}