feat: workflow to polish CLI output with listr2 + chalk

[khaliqgant]

Summary

• Adds workflows/polish-workflow-output.ts — a DAG workflow that integrates listr2 and chalk into the agent-relay workflow output system
• Claude lead plans the integration; two parallel Codex workers implement cli.ts (listr2 step rendering) and runner.ts (chalk color on log/broker/summary)
• Verified with --dry-run (0 errors, 0 warnings)

Test plan

• Run agent-relay run workflows/polish-workflow-output.ts and verify the workflow executes end-to-end
• Confirm cli.ts renders steps with spinners / ✓ / ✗ via listr2
• Confirm runner.ts log output has chalk color on [workflow HH:MM], [broker], and summary table
• Confirm packages/sdk still compiles cleanly after changes

🤖 Generated with Claude Code

---

[devin-ai-integration]

Devin Review found 1 new potential issue.

View 21 additional findings in Devin Review.

[devin-ai-integration]

🟡 Trail left permanently in "active" state when workflow throws

trail start is called unconditionally at line 7, but trail complete/trail abandon at lines 255-261 are only reached if Promise.all on line 10 resolves successfully. The execute() method in packages/sdk/src/workflows/runner.ts:1793-1814 can throw before entering its internal try/catch (e.g., config validation failures, missing workflow name, path validation errors), which would cause Promise.all to reject. Since there is no try/catch around the await Promise.all(...) call, the rejection propagates as an unhandled error, skipping the trail finalization block entirely. The trajectory is left in "active" state indefinitely, polluting the trajectory index.

The fix is to wrap the Promise.all and subsequent logic in a try/catch/finally that ensures trail abandon is called when an exception occurs.

Prompt for agents

In workflows/polish-workflow-output.ts, the trail start at lines 6-8 and the trail complete/abandon at lines 254-261 are not connected by a try/finally, so if Promise.all on line 10 throws, the trail is never closed. Wrap the entire workflow execution (lines 10-262) in a try/catch/finally block so that trail abandon is always called on error. For example:

let result;
try {
  [result] = await Promise.all([
    workflow(...)
      ...run({ onEvent: renderer.onEvent }),
    renderer.start(),
  ]);
  renderer.unmount();
} catch (err) {
  renderer.unmount();
  try {
    execSync('npx trail abandon --reason "Workflow threw an unexpected error"', { stdio: 'ignore' });
  } catch {}
  throw err;
}

try {
  if (result.status === 'completed') {
    execSync('npx trail complete --summary "Polished workflow output with listr2 + chalk" --confidence 0.88', { stdio: 'ignore' });
  } else {
    execSync(`npx trail abandon --reason "Workflow ended with status: ${result.status}"`, { stdio: 'ignore' });
  }
} catch {}

console.log('Result:', result.status);

This ensures the trail is properly abandoned if Promise.all rejects.

---

Was this helpful? React with 👍 or 👎 to provide feedback.