lightspeed-agentic-operator: Skill for Claude Code

Investigate CI Failure

Given a PR URL (e.g. https://github.com/openshift/lightspeed-agentic-operator/pull/123), diagnose why CI jobs failed.

Workflow

1. Extract PR info

Parse org, repo, and PR number from the URL. Fetch metadata with gh:

# PR metadata
gh api repos/{org}/{repo}/pulls/{pr} --jq '{title, state, user: .user.login, head_sha: .head.sha}'

# Changed files
gh api repos/{org}/{repo}/pulls/{pr}/files --jq '.[].filename'

2. Get check statuses

# All checks at a glance
gh pr checks {pr} --repo {org}/{repo}

# Detailed statuses with Prow URLs (use head SHA from step 1)
gh api repos/{org}/{repo}/statuses/{head_sha} \
  --jq '.[] | select(.state == "failure" or .state == "error") | {context, state, target_url}'

This gives you the list of failed jobs and their Prow dashboard URLs.

3. Construct GCS artifact URLs

From a Prow target_url like:

https://prow.ci.openshift.org/view/gs/test-platform-results/pr-logs/pull/{org}_{repo}/{pr}/{job_name}/{build_id}

Derive:

• Directory browser (for navigating artifact tree): https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/test-platform-results/pr-logs/pull/{org}_{repo}/{pr}/{job_name}/{build_id}/
• Raw file content (for fetching logs and JSON): https://storage.googleapis.com/test-platform-results/pr-logs/pull/{org}_{repo}/{pr}/{job_name}/{build_id}/{path}

4. Triage the failure

For each failed job, fetch artifacts in this order:

4a. Quick status

GET storage.googleapis.com/.../finished.json

Check "passed": false and "result": "FAILURE".

4b. Build log (most useful)

GET storage.googleapis.com/.../build-log.txt

This is the main ci-operator build log. It can be large (200KB+). Search from the end for:

• failed / FAILED / error / ERROR
• step .* failed
• Test runner failures (FAIL, --- FAIL, Traceback, AssertionError, Error:)
• Container crash indicators (CrashLoopBackOff, OOMKilled, Error from server)