- Home
- Skills
- DevOps & Cloud
- CI Failure Triage — Find the First Real Failure
More screenshots
Works with the AI Agents you already use
CI Failure Triage — Find the First Real Failure
Reads a failed CI log, separates the first real failure from the cascade after it, and gives a fix order for each failing job.
Free
CI Failure Triage — Find the First Real Failure
Example session with this skill installed
gh run view 123456 --log-failed > failed.log
python3 scripts/ci_triage.py failed.log --out triage.md
build Run npm ci 2026-11-16T10:00:01.000Z npm ERR! code ERESOLVE
unit Run npm test 2026-11-16T10:02:10.000Z AssertionError: expected 900 to be 1000
e2e Run playwright test 2026-11-16T10:03:00.000Z Error: connect ETIMEDOUT 10.0.0.12:5432
- Read your context and instructions
- Compiled the ci failure triage
ci-failure-triage 3 failing jobs 20 error lines
first failure 2026-11-16T10:00:01.000Z build / Run npm ci dependency
npm ERR! code ERESOLVE
root causes 3 cascade 17 score 70/100
CLUSTERS
14 npm ERR! while resolving: react@<n>.<n>
1 npm ERR! code ERESOLVE
1 AssertionError: expected <n> to be <n>
1 Error: connect ETIMEDOUT <n>.<n>:<n>
FIX ORDER
- build / Run npm ci [dependency]
Reproduce the install locally with the committed lockfile, then fix the manifest and lockfile together. - unit / Run npm test [test]
Open the first failing assertion, reproduce it locally, and fix the code or the expectation. - e2e / Run playwright test [infrastructure]
Re-run the failed job once; if it reproduces, fix the runner, network or resource limit.
HIGH infrastructure e2e / Run playwright test
Error: connect ETIMEDOUT 10.0.0.12:5432
Connects securely to your tools. The creator never sees your data.
About this skill
A red workflow prints one cause and then dozens of consequences. The dependency install fails, the test job reports a missing module, the end-to-end job times out, and the last error in the log is usually the least useful one. This finds the first real failure and separates it from the noise.
It accepts the log a developer already has:
gh run view --log-failedoutput, with job, step and timestamp- GitHub Actions raw logs with
##[group]Run …markers - A plain text log with one failure per line
Then it reports:
- The first failure in each job, with the step it came from
- One root-cause candidate per failing job, not the last error printed
- The cascade count: how many error lines follow the first failure
- Clusters of repeated messages, normalized across paths, hashes and numbers
- A fix order with one action per failure class
Each root cause is labelled dependency, infrastructure, authentication, build, test or unknown. Forty copies of the same error become one cluster line with a count.
The shipped failing log is three jobs and 20 error lines. The first failure is npm ci with a peer-dependency conflict, followed by 17 cascade lines. A unit assertion and an e2e connection timeout are separate root causes. Score 70/100, exit 1. The clean log exits 0.
Good to know: classification is a heuristic over text, and unknown is a valid answer. The earliest failure is the best candidate in the log, not a proven root cause. Truncated logs, redacted secrets and missing timestamps reduce accuracy. The tool never calls GitHub, downloads logs, reruns a job or changes a workflow. It reads a file you already have, and it says so when the format gives it no timestamps.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
Classification uses heuristics and may return unknown. Accuracy depends on log completeness and timestamps. Does not call GitHub APIs or rerun jobs.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 day ago
- Free with an Agensi account