fix(terminal): settling a thread closes its idle shells

[t3dotgg]

Settled threads kept their terminal shells open forever. The worktree setup shell stays at a prompt after its script ends. Also, zsh prompt themes (zsh-async and similar) fork a helper copy of the shell, and the terminal manager counted that copy as a running command. So every idle zsh shell showed the pulsing terminal icon in the sidebar, even on settled threads. A live terminal also blocks worktree cleanup for its checkout.

Earlier attempts stopped at "never close terminals on settle" (#5774, #4684) because closing every terminal can kill a dev server or an editor. This change closes only shells that wait at an idle prompt:

• Idle detection: a child process that is a copy of the shell and has no children of its own is not a running command.
• TerminalManager.closeIdle: closes a thread's terminals that wait at an idle prompt. It runs a fresh process check and does not use the last poll. A terminal that runs a command stays open. If the check fails, all terminals stay open.
• Settle: the existing thread.settled hook in the provider command reactor calls closeIdle. That covers manual settle and auto-settle.
• Setup script: after the script exits 0, the runner closes its setup shell if it is idle. A failed script keeps its shell open so the user can look at the error.

Closed terminals keep their output in the terminal history. A reopened terminal starts a new shell. No wire contract changed, so web, desktop, and mobile all get this from the server.

Proof: I ran the real TerminalManager with real node-pty zsh shells and ps. On main, an idle shell reported running: true (zsh) because of its prompt worker. On this branch, it reports false. closeIdle closed the idle shell and kept the shell that ran sleep 60. Focused tests cover detection, closeIdle, the settle hook, and the setup runner.

Made with Claude Opus 5.5 (1M context) in Claude Code, running in T3 Code.

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features
  • Settling a thread now closes its idle terminals while preserving their output. Terminals running commands remain open.
  • Setup shells close after a successful script exit when idle; shells remain open after unsuccessful or incomplete runs.
• Documentation
  • Updated thread guidance to explain which terminals close when a thread is settled.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

---

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

[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	+10 B (+0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	+9 B (+0.1%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.3 KiB	56.3 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+10 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	−1 B (−0.0%)	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: fd996d1 · PR result: 25bd7cd · 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: 114.0 KiB
• Claude decoded thread snapshot: 114.7 KiB

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

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR changes production terminal lifecycles by closing idle shells during settlement and successful setup completion, using new process-tree detection and race-handling logic in shared terminal infrastructure. The behavior is intentionally narrow and tested, but its observable side effects and cross-component runtime impact merit human review.

No code changes detected at 25bd7cd. Prior analysis still applies.

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