Summary

Reproduce startup behavior when the GitHub CLI (gh) is entirely absent or not on the PATH; add unit tests for HealthService, implement an explicit ENOENT detection and a small non-blocking retry/backoff so the server recovers when gh becomes available, and add a short runbook/docs snippet with reproduction and remediation steps.
(see parent issue #7687)

To do

• Add reproducible instructions and a short script to simulate gh being absent (e.g., PATH modification) under ai/mcp/server/github-workflow/docs/gh-absent.md.
• Add unit tests for HealthService that mock child_process.exec / execAsync to simulate these states:
  • ENOENT (gh not installed / not in PATH)
  • stdout empty + stderr-only output
  • gh present but unauthenticated
  • gh present but old version
• Implement code changes in ai/mcp/server/github-workflow/services/HealthService.mjs to:
  • Detect ENOENT explicitly and log a clear, actionable message ("gh CLI not found in PATH; tools will be disabled until installed")
  • Return a consistent health status (degraded/unhealthy) without crashing the server
  • Add a small non-blocking retry/backoff (configurable) for auth/version checks so the server can recover without restart
• Add a short runbook/docs entry with reproduction steps and remediation commands.

Acceptance criteria

• Reproduction steps and logs included in the issue or added doc file
• Unit tests added covering the four states; tests pass locally and in CI
• HealthService code updated with ENOENT handling and retry/backoff and no server crash on startup
• PR references this issue and parent Enhance GitHub Workflow server robustness when gh cli is not installed #7687