- Home
- Skills
- DevOps & Cloud
- Failed CI Run Triage
Works with the AI Agents you already use
Failed CI Run Triage
CI failure triage for GitHub Actions, GitLab and Docker builds: finds the first real cause in a failed build log. Local and read-only.
$5
Failed CI Run Triage
Example session with this skill installed
/ci-failure-triage ci-log.txt
- Read your context and instructions
- Compiled the failed ci run
Illustrative. Both logs are invented (not from a real project); the reports below are what the script prints.
Log 1: a container build that ran out of memory
CI FAILURE TRIAGE (12 lines read)
Failing step: [build 4/6] RUN npm run build
Most likely first cause: Out of memory (line 5, 1 matching line(s))
evidence: #9 45.11 FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory
next: Look at the memory the job had. Re-run once; if it repeats, raise the runner memory or lower parallelism.
All findings, in the order they appear:
- line 5 Out of memory x1 [[build 4/6] RUN npm run build]
#9 45.11 FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory
- line 11 Container build x1 [[build 4/6] RUN npm run build]
ERROR: failed to solve: process "/bin/sh -c npm run build" did not complete successfully: exit code: 134
Later findings (Container build) come after the first one and may be side effects of it. Fix the first, then re-run.
Lines that only say the step failed:
- line 12: ##[error]Process completed with exit code 1.
Limits: pattern matching on text. It cannot see what is missing from a truncated log, and a flaky failure and a real one can look the same. Credentials found in lines are masked.
Log 2: a test run that failed on a network error
CI FAILURE TRIAGE (10 lines read)
Failing step: Run npm test
Most likely first cause: Network or external service (line 7, 1 matching line(s))
evidence: FetchError: request to api.example.test/v1/profile failed, reason: read ECONNRESET
next: Check the status page of the service named in the line, then re-run once. Repeated failures point to a blocked host or a missing proxy setting.
All findings, in the order they appear:
- line 5 Test failure x2 [Run npm test]
FAIL src/api.test.js
- line 7 Network or external service x1 [Run npm test]
FetchError: request to api.example.test/v1/profile failed, reason: read ECONNRESET
A test failure is also present. If the cause above is an outside condition (network, timeout, memory), the test may have failed because of it: re-run once before changing code.
Lines that only say the step failed:
- line 9: npm ERR! code ELIFECYCLE
- line 10: ##[error]Process completed with exit code 1.
Limits: pattern matching on text. It cannot see what is missing from a truncated log, and a flaky failure and a real one can look the same. Credentials found in lines are masked.
Next steps
- Read the first finding, not the last line: the red exit code is only the symptom.
- Re-run once when the first cause is an outside condition (network, timeout, memory) before changing code.
Connects securely to your tools. The creator never sees your data.
About this skill
Point it at a failed CI log and get the first real cause instead of thousands of lines of noise. The script finds the failing step, sorts what it sees into known kinds (test failure, compile or type error, dependency install, out of memory, disk full, timeout, network or external service, permission, missing variable, container build, lint), says which finding came first, notes which later findings may be side effects, and gives the next check for that kind of problem. It understands GitHub Actions, GitLab, Docker BuildKit and plain text logs.
What it reads and writes: the one log file you name; it writes nothing. If you paste a log into the chat, the skill first saves it as ci-log.txt in your current folder and tells you. No network connection, no environment variables, nothing to install. Tokens, bearer headers and passwords found in printed lines are masked.
Honest limits: it is pattern matching on text. A truncated log hides what is missing, a flaky failure and a real one can look the same in a single log, and it knows ten kinds of failure, not every tool. It does not re-run your pipeline and changes no code. Needs Python 3.9+.
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.
Needs Python 3.9+. Text pattern matching only: a truncated log hides causes, a flaky and a real failure can look the same, and it knows ten kinds of failure. Credentials are masked in printed lines, but check before sharing a report.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 15 days ago
- Software
Python
Permissions
File Scopes
Reads only the log file you name; no network connection, no environment variables, no other files. When you paste a log into the chat, the skill first saves it as ci-log.txt in the current folder. Words such as token or password appear in the code as patterns used to mask secrets in printed lines.