Back to Skills Marketplace
Skill pack

CI/CD Failure Triage

Investigate failing pipelines, workflow logs, flaky tests, failed deploys, release blockers, and safe rerun recommendations.

v0.1.0 1 skill tools By Kendr cicdgithub-actionspipelinestestsdeploymentsdevops
Publisher docsHosted archive

How to use it

  1. Open Kendr Desktop.
  2. Go to Skills, then Marketplace.
  3. Search for CI/CD Failure Triage or cicd-failure-triage.
  4. Install the pack, then enable the pack or individual skills you want available in agentic mode.

Kendr Desktop compares the installed version with the hosted catalog version and offers an update when this pack changes.

Install source

Use this hosted archive when installing or updating the pack from Kendr Desktop.

https://kendr.org/api/skills/packs/cicd-failure-triage/archive

CI/CD Failure Triage

Triage CI/CD failures using GitHub, repository context, logs, tests, and deployment history before proposing safe reruns or fixes.

CI/CD Failure Triage

Use this skill when the user asks why a build failed, why a deploy is blocked, whether a workflow can be rerun, or how to summarize failed CI/CD runs.

Expected companion packs:

  • GitHub Official MCP for repositories, pull requests, workflow runs, jobs, and logs.
  • Docker Operations when failures involve local container builds, images, Compose, or runtime checks.
  • Grafana Official MCP when failures correlate with production alerts or deployment impact.

Workflow:

  • Identify repository, branch, pull request, commit SHA, workflow run, failed job, and failing step.
  • Read logs before proposing fixes. Extract the first root-cause error, not just the final command failure.
  • Compare recent successful runs when possible to distinguish new regressions from environment failures.
  • Classify the failure as test regression, dependency/cache issue, infrastructure issue, auth/secret issue, flaky test, deployment guardrail, or unknown.
  • Return a concise triage with root cause, evidence, affected workflow, likely owner, and next action.

Actions:

  • Treat rerun jobs, cancel workflow, approve deployment, merge, release, tag, or commit changes as side-effecting.
  • Use Kendr approval before side-effecting GitHub or deployment actions.
  • Prefer rerunning only failed jobs when the evidence indicates flakiness or transient infrastructure failure.
  • Do not expose secrets from logs. Redact tokens, credentials, and private URLs.