[waleedlatif1]

Summary

• A failed shell command's error was its raw stderr, and the exit code appeared only when stderr was empty. A process killed by a signal (out of memory, a terminated run) that had printed a harmless warning first reported that warning as the error, so the agent chased the warning (e.g. "set an HF_TOKEN") instead of the kill
• About a fifth of failed sandbox shell commands in a day of production logs were signal kills (mostly SIGTERM, then SIGKILL)
• Shell errors now always lead with the exit code, naming the signal for a 128 + n code, followed by the process's own output: Process exited with code 137 (the exit code of SIGKILL, usually out of memory): <stderr>. The code alone cannot prove a kill (a program may exit 137 itself), so it is described as the signal's code, not asserted
• Python/JavaScript failures keep the headline they printed as the error name and value; for a 128 + n exit code the same exit description is appended (<value> [Process exited with code 137 (…)]), so neither a program's own error nor the exit code is lost. Other exit codes are unchanged
• Shared by Mothership run_code and workflow Function blocks; nothing parses the message

Type of Change

• Bug fix

Testing

• lib/execution/remote-sandbox/conformance.test.ts (E2B and Daytona): a command killed with 137 after printing a warning reports the signal first and keeps the warning; fails on staging for both providers. The existing "real command output" case still holds
• lib/execution/remote-sandbox/code-failure.test.ts: a printed headline survives a 137 exit with the exit description appended, and exit code 1 is unchanged; exit-code descriptions at 1, 128, 129, 143 and 193
• 219/219 in both files; bun run lint, check:utils, check:comment-hygiene, check:test-patterns, check:explicit-any, check:source-conventions
• Type-check, full check:audits and the test suites run 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 10:01am UTC

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge.

Summary

The PR adds exit-code descriptions to sandbox errors while keeping the process’s printed output.

• Failed sandbox runs show exit codes beside their printed output.

_{Reviews (2) · Last reviewed commit: "fix(execution): describe a signal's exit..." · 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 4 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}