[waleedlatif1]

Summary

Follow-ups to #9507 found in the release audit:

• Stop cancelled settled and paused runs. getDetachedWorkflowExecutions returned every execution a call in the stream ever detached, and Stop wrote the durable 14-day cancel flag for each. A run paused at a human-in-the-loop block resumes under the same execution id, read the flag, and was cancelled. Stop now cancels only runs whose log reads running (or that have no log yet, i.e. still starting); paused runs are left alone, matching the pre-feat(mothership): answer a long workflow tool run still running, and stop detached runs on Stop #9507 behaviour where a paused browser run was already settled
• A held workflow run blocked the next turn's workbench. answerWithinWait retains the run so the call's executionSettledAt stays null while it runs; prepareWorkbenchAccess counted that as handlersPending, so every run_code/sandbox call in the next turn failed for up to 95 minutes. Workflow-tool rows (which never touch the chat's workbench) no longer count; their unsettled sandbox processes still feed recovery
• Stop read full JSONB payloads inside transactions. The status/exit-code filters now run in SQL and only the argv and status fields are selected, so member/workspace/chat/account deletion no longer materializes every sim_cli stdout of a stopped turn
• Reuse WORKFLOW_TOOL_NAMES instead of a duplicate list; explicit settled-state tracking in answerWithinWait instead of a Promise.race ordering trick; the admission log poll is capped at 60 s past the wait

Type of Change

• Bug fix

Testing

• workflow-client-settlement.integration.ts (real Postgres + Redis): detached CLI and tool runs whose log is pending/completed/failed/cancelled are not cancelled; a run answered still running with no log yet still is; a later turn's workbench access isn't blocked by an earlier still-running run_workflow, while an earlier function_execute still blocks (control). Ran locally: 21/21; with the pre-fix repository 5 fail
• Unit tests in tools/handlers/workflow, async-runs, request/application, request/handlers (209 pass); biome; full gate in CI

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing (new tests pass the test-audit authoring gate)
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Oct 31, 2026 6:21pm UTC

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable new issues were found.

Summary

This PR narrows Stop to detached workflow runs that are still running and lets a later turn use the workbench while an earlier workflow run continues.

• Stop cancels detached workflow runs only while they are running.
• A running workflow tool no longer blocks the next turn’s workbench.
• Workflow tools return settled results when runs finish near the wait limit.

No actionable new issues were found. Tests were inspected, not run.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Stop] --> B[Find runs started by this turn]
  B --> C{Current log status}
  C -->|Running or no log| D[Cancel execution]
  C -->|Paused or finished| E[Leave execution alone]
  F[Later turn requests workbench] --> G[Check earlier tools and sandbox processes]
  G --> H[Workflow handler alone does not block]
  G --> I[Unfinished sandbox processes still require recovery]

Loading

_{Reviews (2) · Last reviewed commit: "fix(mothership): answer a workflow tool ..." · Reviewed by Greptile}

[cubic-dev-ai]

All reported issues were addressed across 4 files

<sub>Reply with feedback, questions, or to request a fix.

Turn on auto-fix | Re-trigger cubic</sub>

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 6 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Turn on auto-fix | Re-trigger cubic}