Fix: Buildkite detective skip conditions exit cleanly instead of failing

[Copilot]

The buildkite PR detective workflow was reporting failures when skip conditions were met (no PR association, build not failed, no failed jobs). The skip() function used sys.exit(1), causing the job to fail rather than complete successfully.

Changes

• .github/workflows/gh-aw-estc-pr-buildkite-detective.md: Changed skip() function exit code from 1 to 0
• .github/workflows/gh-aw-estc-pr-buildkite-detective.lock.yml: Recompiled workflow with updated exit code

def skip(reason):
    subprocess.run(['bash', '-c', f'echo "::notice::{reason}"'], check=False)
    print(reason)
    sys.exit(0)  # was: sys.exit(1)

The workflow still emits notice messages but now completes successfully when skipping, preventing false failure notifications.

• Fixes skip exits with 1 for the buildkite pr build analyser #1946

[v1v]

@copilot run the make compile

[v1v]

@copilot given https://github.github.com/gh-aw/ can we configure the skip for the other jobs to be skipped conditionally somehow?

[Copilot]

@copilot run the make compile

Done! I've run make compile and committed the updated lock files. The compilation applied compiler improvements including fixing report-failure-as-issue to use the input parameter instead of being hardcoded. Changes committed in c771c9b.

[Copilot]

@copilot given https://github.github.com/gh-aw/ can we configure the skip for the other jobs to be skipped conditionally somehow?

Yes, there are two approaches to conditionally skip the agent job when there's no data:

Option 1: Move data fetch to activation job (better architecture)
Move the "Resolve event context and fetch Buildkite data" step from the agent job to the activation job, set a step output (e.g., has_data: 'true'/'false'), then make the agent job conditional:

agent:
  needs: activation
  if: needs.activation.outputs.has_data == 'true'

Option 2: Use noop safe output (current architecture)
Keep the current structure but instead of calling skip() early, let the agent run and emit a noop safe output with "No data to analyze". The workflow already has safe-outputs: noop: configured for this pattern.

Option 1 is cleaner because it avoids starting the agent job entirely when there's nothing to do. The current fix (exit 0) prevents false failures, but the agent still runs and consumes resources. Would you like me to implement Option 1?

[github-actions]

TL;DR

Internal: CI failed in Verify compiled workflows because generated lock files were stale after source workflow edits. Re-run make compile and commit the resulting .lock.yml updates from this branch.

Remediation

• Run make compile locally in this PR branch and commit all changed generated lock files.
• Push that commit; CI should pass once Ensure lock files are up to date sees a clean diff.

Investigation details

Root Cause

The failing step is a generated-artifact consistency check, not a runtime or logic failure. Source workflow changes landed without the corresponding regenerated lock files, so the CI diff check fails.

Evidence

• Workflow: https://github.com/elastic/ai-github-actions/actions/runs/33170294536
• Job/step: Verify compiled workflows → Ensure lock files are up to date
• Key log context: all other jobs succeeded; only the lock-file freshness gate failed immediately after compile.

Validation

• Not run in this environment (read-only detective workflow).

Follow-up

• If this recurs often, consider adding a local pre-push/pre-commit reminder to run make compile when touching gh-aw-*.md sources.

---

What is this? | From workflow: Trigger PR Actions Detective

Give us feedback! React with 🚀 if perfect, 👍 if helpful, 👎 if not.