Two Copilot CLI workflows failed with node: command not found in a 1.5-hour burst on 2026-04-27, and one Codex workflow failed with codex: command not found in the same 6h window. Both failure modes share the same root cause: the engine binary was absent from PATH when the runner agent step launched, killing the workflow before any agent output was produced.

Affected Workflows

Root Cause

Runner slot(s) provisioned without the engine binary in PATH during the 09:25–11:41 UTC window. The Copilot CLI is Node.js-based; when node is absent the CLI fails immediately at startup. The Codex binary is a separate installation that similarly was not found.

This is intermittent — other Copilot CLI runs (Auto-Triage Issues, PR Triage Agent, Visual Regression Checker, Smoke CI, Delight) and other Codex runs (AI Moderator ×2, Duplicate Code Detector) all succeeded in the same 6h window, confirming correct provisioning on most runner slots.

Proposed Remediation

1. Verify that the runner setup action for each engine includes a post-setup node --version / codex --version health-check that fails fast with a clear diagnostic before the agent job starts.
2. Add a setup-node step (or ensure actions/setup-node runs) in the Copilot CLI agent job to guarantee Node.js is in PATH before the CLI is invoked.
3. Investigate whether a specific runner label or pool is responsible for slots that intermittently lack the required binary.

Success Criteria

Zero recurrence of command not found engine startup errors across 5+ consecutive scheduled runs of affected workflows.

Parent: #28673

References:

• §24986870660 — Daily News (failure)
• §24990655972 — Daily Issues Report Generator (failure)
• §24992928191 — Daily Fact About gh-aw (failure)

Generated by [aw] Failure Investigator (6h) · ● 631.2K · ◷

• expires on May 4, 2026, 1:25 PM UTC