Skip to content

Harden DAG task runner startup and env handling#33

Draft
cursor[bot] wants to merge 2 commits intomainfrom
cursor/critical-bug-inspection-a84f
Draft

Harden DAG task runner startup and env handling#33
cursor[bot] wants to merge 2 commits intomainfrom
cursor/critical-bug-inspection-a84f

Conversation

@cursor
Copy link
Copy Markdown

@cursor cursor Bot commented May 10, 2026

Summary

  • Keep Agent.create() and initial agent.send() inside the per-task timeout/error path so a stalled SDK setup marks only that task ERROR instead of leaving the DAG stuck in RUNNING.
  • Preserve final run.wait().result as a capped fallback when no assistant stream text was captured, so downstream tasks receive parent output.
  • Remove skill instructions that auto-source .env, avoiding arbitrary shell execution from workspace files.

Validation

  • pnpm typecheck in sdk/dag-task-runner
  • pnpm install && pnpm typecheck in .cursor/skills/dag-task-runner/scripts
  • pnpm init-canvas in sdk/dag-task-runner
Open in Web View Automation 

cursoragent and others added 2 commits May 10, 2026 11:09
Co-authored-by: Jon Kaplan <jkap-anysphere@users.noreply.github.com>
Co-authored-by: Jon Kaplan <jkap-anysphere@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant