[waleedlatif1]

Summary

• When run_block, run_from_block, run_workflow, or run_workflow_until_block ran in the browser and the workflow failed, the tool error the agent read was always the generic Workflow execution failed.. The real message (<Block>: <reason>) only showed up nested in output.output.error.
• Cause: getTrustedWorkflowToolExecution read only the top-level executionData.error. Error finalization (completeWithError and its cost-only fallback, finalizationPath: 'force_failed') never writes that field. It stores the message as finalOutput: { error }.
• Fix: the trusted read now takes finalOutput.error when the run was error-finalized (force_failed). This string has already gone through filterForDisplay and PII redaction, and the agent already sees it nested in output, so nothing new is exposed.
• A run that failed only through its block spans keeps its workflow output, so its finalOutput.error is workflow data. That case is not read. Cancelled runs are unchanged.
• Effect on consumers: in client.ts the message now becomes the top-level tool error and the persisted tool-call row error, and it is added to the trusted provenance import. It still passes through boundRunResultForModel and the secret-provenance projection unchanged. /api/copilot/confirm reads only status and executionId, so it is not affected.

Type of Change

• Bug fix

Testing

• New trusted-tool-failure.integration.ts runs against real Postgres. It finalizes runs through a real LoggingSession and reads them back through getTrustedWorkflowToolExecution.
  • Thrown run returns its message. This test fails on the pre-fix code.
  • Run that failed through its block spans with a workflow { error } output returns no message. This test fails when the force_failed gate is removed.
• execution-state.test.ts, mothership/request/tools/client.test.ts, and copilot/confirm/route.test.ts pass (118 tests). bun run lint passes.

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		Nov 3, 2026 9:03pm UTC

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; the cleanup change fixes the earlier finding without introducing a new issue.

Summary

Browser-settled workflow tools now read the saved failure message when finalizationPath is force_failed. Other runs keep treating finalOutput.error as workflow data.

• Browser-run workflow tools show the saved reason when a run fails.

_{Reviews (2) · Last reviewed commit: "fix(mothership): clean up run logs and s..." · Reviewed by Greptile}

[cubic-dev-ai]

All reported issues were addressed across 2 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 2 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}