[ThomasK33]

Summary

Moves the stream-end decisions out of TaskService into a new plain class, StreamEndLedger (src/node/services/streamEndLedger.ts). The move is move-only. taskService.ts drops from 20,307 to 20,023 lines. PR 10, the last PR of the TaskService stop-and-attempt split, on top of #7678.

Across this stack (#7674, #7675, #7677, #7678 and this PR), taskService.ts goes from 21,742 to 20,023 lines. The stop and attempt state now lives in three owner classes: TaskStopLatch, AttemptLedger and StreamEndLedger.

Implementation

• Fields that moved: streamEndDecisionsByTaskId (readonly, because tests read it) and deferredTaskStreamEnds. The inline value type of deferredTaskStreamEnds is now the named interface TaskStreamEndDeferral, with the same fields and comments.
• Methods that moved, with names and statement order kept: runStreamEndHandler, deferChildStreamEnd, registerStreamEndDecision, findStreamEndDecision, findPendingStreamEndDecision, resolveStreamEndDecision, pruneStreamEndDecisions, reconcileDeferredTaskStreamEnd and resolveDecisionOfReportedAttempt.
• Coordinators stay: recordWorkspaceTurnSettled and recordWorkspaceTurnSuperseded stay on TaskService. Their decision loops become one settleTurn / rebindTurn call each, at the same position, so the order across the three owners does not change.
• Same-tick order: the stream-end listener in the constructor keeps its one-tick order (snapshot, origin, register, run).
• Stays on TaskService: handleStreamEnd, inspectAttemptOutcome and reawakenInterruptedTask. A later plan covers these long methods.
• Cross-class calls: the AttemptLedger and TaskStopLatch closures findStreamEndDecision, pruneStreamEndDecisions and discardStreamEndDecisions now point at streamEndLedger through TaskService. The three classes never call each other directly.
• Facades: none. No test spies on, patches or casts a moved method.

Validation

• The new file streamEndLedger.test.ts has 21 tests on fake deps. They cover:
  • decision identity
  • find order
  • the order of resolve, prune, stop recheck and drain
  • the prune rules
  • settle and rebind of turns
  • handler outcomes (a return gives nonreport, a throw gives indeterminate)
  • the deferral listener
  • each branch of reconcileDeferredTaskStreamEnd
• Three TaskService test files changed only their cast targets. Prettier re-wrapped some of the longer lines.
• CRAP: every moved function scores CRAP 13 or less with coverage. The highest is reconcileDeferredTaskStreamEnd at CC 13 with 100 % line coverage.

Risks

Medium-low. The stream-end decision says whether a child's last stream ended in a report. A wiring mistake would hold or drop a report, or hold a stop latch. The report-decision, settlement-receipt and child-goal suites pass, and they changed only their cast targets.

---

Generated with xum • Model: anthropic:claude-opus-5-5 • Thinking: high • Cost: $68.74

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-27T13:17:59.911089Z	64d7681	Manual request
🔒 Security Review	✅ Completed 2026-10-27T13:19:48.692721Z	64d7681	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[ThomasK33]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Breezy!

Reviewed commit: 64d7681e57

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

🛡️ Codex Security Review · Automatically triggered

Security review completed. No security issues were found in this pull request.

Reviewed commit: 64d7681e57

View security finding report

Only the user who started this review can view the report in Codex.

ℹ️ About Codex security reviews in GitHub

This is an experimental Codex feature. Security reviews are triggered when:

• You comment "@codex security review"
• A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review

Once complete, Codex will leave suggestions, or a comment if no findings are found.

[github-actions]

🟢 e2e stories (Haiku): 14 passed

22 agent steps · 7 replayed from cache · 15 model calls · 46.2k tokens (24% cached)
Spend: $0.09 of the $1.00 cap (proxy ledger)

All 14 tests in 9 files

	Test	Agent	Time
🟢	chatRestart.e2e.ts · 1 passed	1 step · 1 call	6.1s
🟢	a chat survives a restart of the server	1 step · 1 call	6.1s
🟢	compaction.e2e.ts · 1 passed	1 step · 1 call	4.8s
🟢	/compact replaces the old chat with a summary	1 step · 1 call	4.8s
🟢	pluginTrustRestart.e2e.ts · 1 passed	4 steps · 2 calls	10.9s
🟢	a trusted plugin hook stays trusted after the server restarts	4 steps · 2 calls	10.9s
🟢	settingsRoundTrip.e2e.ts · 1 passed	2 steps · 1 call	9.7s
🟢	a chat survives trips through Settings	2 steps · 1 call	9.7s
🟢	sidebarPerWorkspace.e2e.ts · 1 passed	4 steps · 1 call	5.6s
🟢	a new workspace starts with its right sidebar collapsed while another keeps it open	4 steps · 1 call	5.6s
🟢	smoke.e2e.ts · 1 passed	2 steps · 1 call	3.0s
🟢	the agent opens the playground workspace	2 steps · 1 call	3.0s
🟢	streamErrors.e2e.ts · 1 passed	1 step · 1 call	6.9s
🟢	stream errors show in the chat, and the chat recovers	1 step · 1 call	6.9s
🟢	todoNudge.e2e.ts · 1 passed	1 step · 1 call	4.1s
🟢	the agent finishes open TODO items after the turn-end nudge	1 step · 1 call	4.1s
🟢	touchAlignment.e2e.ts · 6 passed	6 steps · 6 calls	23.9s
🟢	the Settings close button is level with the title on a touch screen 390px wide	1 step · 1 call	4.7s
🟢	the Settings close button is level with the title on a touch screen 768px wide	1 step · 1 call	4.4s
🟢	Backup settings checkboxes are small and level with their labels on a touch phone	1 step · 1 call	4.6s
🟢	the heartbeat checkbox is small and level with its label on a touch phone	1 step · 1 call	3.4s
🟢	a reply's model and time stay on one line on a touch phone	1 step · 1 call	3.6s
🟢	the workspace actions menu stays on the title row on a touch phone	1 step · 1 call	3.2s

_{e2e 0.20.0 · 1m 19s · web · run artifacts}

🟢 e2e Electron stories (Haiku): 3 passed

3 agent steps · 3 model calls · 7.9k tokens (22% cached)
Spend: $0.02 of the $1.00 cap (proxy ledger)

All 3 tests in 3 files

	Test	Agent	Time
🟢	electron/deepLink.e2e.ts · 1 passed	1 step · 1 call	6.9s
🟢	a deep link of a second launch starts a chat in the running app	1 step · 1 call	6.9s
🟢	electron/menuBar.e2e.ts · 1 passed	1 step · 1 call	8.4s
🟢	the workspace menu bar is the desktop title bar	1 step · 1 call	8.4s
🟢	electron/smoke.e2e.ts · 1 passed	1 step · 1 call	6.2s
🟢	Xum Electron opens the playground workspace	1 step · 1 call	6.2s

_{e2e 0.20.0 · 24.2s · electron · run artifacts}

🟢 e2e agent repros (Haiku): no tests selected

Spend: $0.00 of the $1.00 cap (proxy ledger)

_{e2e 0.20.0 · 7.1s · 2 targets (web, phone) · run artifacts}