Fetch complete GitHub Actions job logs and diagnose failed CI jobs from saved evidence.
npx skills add https://github.com/radixark/miles --skill ci-fetch-log
Please help me install this Agent Skill. Skill: ci-fetch-log Source: https://github.com/radixark/miles/tree/main/.claude/skills/ci-fetch-log Command: npx skills add https://github.com/radixark/miles --skill ci-fetch-log
CI failures on GitHub Actions are hard to investigate when logs are viewed piecemeal in the browser, pasted as snippets, or lost after reruns. This Skill downloads complete job logs via the gh CLI, writes an auditable completeness manifest, and optionally diagnoses failed jobs strictly from the saved evidence.
gh api or gh run view, and records per-job completeness in a machine-readable manifest.json.report.md files.run.json, manifest.json, and summary.md in a timestamped directory outside the repository so results can be reviewed or consumed by other tools later.Fetch the complete logs for the latest CI run on PR 482 and diagnose any failed jobs from the saved evidence.
High-intent search queries and answers about installing and using this skill.
Use the gh CLI: run `gh api /repos/<owner>/<repo>/actions/jobs/<job_id>/logs` to save a job's full log, with `gh run view <run_id> --log --job <job_id>` as a fallback. This Skill automates both commands and records which succeeded per job.
Resolve the PR head SHA with `gh pr view <number> --json headRefOid`, then list runs for that commit with `gh run list --commit <sha>`. The Skill picks the newest run by createdAt and databaseId and never falls back to an older commit.
Yes, but single-job selection requires the `--only` flag together with a job URL, an explicit `--job <id>`, or a `--job-name <glob>` selector. Without `--only`, the Skill always selects the whole run to guarantee completeness.
Authentication fails when GH_TOKEN or GITHUB_TOKEN is missing, GH_CONFIG_DIR points elsewhere, or shell startup skips credential setup. Run `gh auth status --hostname github.com` in the exact shell that will execute fetches; the Skill stops before any fetch if this gate fails.
Diagnosis only covers jobs whose saved log is verified complete; queued, running, or partially fetched jobs are listed separately and never diagnosed from partial evidence. Every claim must cite exact line spans from the saved log, and weak evidence yields an 'instrument first, fix later' recommendation.
euxxDebug failing GitHub Actions checks by locating runs, reading logs, and summarizing root causes.
AgentSystemLabsDiagnoses GitHub CI test failures and applies minimal fixes to resolve them.
packetalienAnalyze CI run logs and test histories to detect flaky tests.
kennyolofsson23-netizenInspect failing GitHub Actions checks and extract failure context from PR runs.