gh-run-failure
用于分析GitHub管道或作业中的失败。
name: gh-run-failure description: Use to analyze failures in GitHub pipelines or jobs.
- Assume the `gh` tool is installed and configured. - Avoid printing full logs; focus on the failing step. - For large logs, prefer downloading artifacts. - If a test case failed, specify which test and how to run it locally.List recent runs
gh run list --limit 20 --json databaseId,createdAt,headBranch,event,conclusion,status,displayTitle \ --template '{{range .}}{{printf "%8v %-20s %-8s %-10s %-10s %s\n" .databaseId .createdAt .status .conclusion .event .displayTitle}}{{end}}'Inspect a single run
RUN_ID=123456789 gh run view "$RUN_ID" --json status,conclusion,createdAt,headSha,headBranch,event,workflowName,urlShow only failed jobs or steps
gh run view "$RUN_ID" --json jobs --jq ' .jobs[] | select(.conclusion != "success") | .name as $job | (.steps[]? | select(.conclusion != "success") | "\($job) :: \(.name) :: \(.conclusion)")'View failed jobs/steps output
gh run view "$RUN_ID" --log-failedTo focus on a specific job, first list them: gh run view "$RUN_ID" --json jobs --jq '.jobs[] | "(.databaseId)\t(.name)\t(.conclusion)"'
Then fetch that job's log: JOB_ID=987654321 gh run view "$RUN_ID" --job "$JOB_ID" --log
List artifacts from a run
gh run download "$RUN_ID" --listDownload a specific artifact
gh run download "$RUN_ID" --name "test-logs" --dir /tmp/github-run-"$RUN_ID"After downloading, search locally for errors: rg -n "error|fail|exception" /tmp/github-run-"$RUN_ID"
Compare failed and latest successful run
Get latest successful run ID:
BRANCH=main
OK=$(gh run list --branch "$BRANCH" --limit 50 --json databaseId,conclusion
--jq '.[] | select(.conclusion=="success") | .databaseId' | head -n1)
echo "$OK"
Diff their SHAs: FAIL_SHA=$(gh run view "$RUN_ID" --json headSha --jq .headSha) OK_SHA=$(gh run view "$OK" --json headSha --jq .headSha) gh api repos/:owner/:repo/compare/"$OK_SHA"..."$FAIL_SHA" --jq '.commits[].commit.message' | sed 's/^/- /'