fix-ci

Diagnose failing CI checks and apply the smallest verified fix.

Updated Sep 3, 2026
One-click install
npx skills add https://github.com/JonusNattapong/A2A-MCP --skill fix-ci-jonusnattapong
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fix-ci
Source: https://github.com/JonusNattapong/A2A-MCP/tree/main/src/agenttalk/skills/devkit/fix-ci
Command: npx skills add https://github.com/JonusNattapong/A2A-MCP --skill fix-ci-jonusnattapong

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? When a test, lint, build, security, or CI job goes red, it is easy to guess at causes, over-fix, or widen scope into unrelated work. This Skill enforces a disciplined loop: read the full failing log first, classify the root cause, apply the smallest fix, and verify the check is green again. ## Core Features & Use Cases - Log-first diagnosis: Captures the exact failing command, ref, environment, and full output before any editing, and reproduces locally when deterministic and cheap. - Root-cause classification: Sorts failures into code defect, test defect, flaky test, environment, dependency, or CI configuration before choosing a fix. - Smallest-fix discipline: Applies or proposes only the change that addresses the root cause, explicitly avoiding drive-by refactors, scope creep, and self-approval. - Use Case: A CI lint job fails on a pull request. The Skill reads the CI log, identifies a dependency version mismatch as the root cause, pins the dependency, reruns the original command to confirm green, and emits a production-handoff record for reviewers. ## Quick Start Ask the agent to diagnose the failing CI job, find its root cause from the logs, and apply the smallest fix that makes the check pass again.

Frequently Asked Questions about fix-ci

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fix a failing CI check?▼

Start by capturing the exact failing command, ref, environment, and full log output before editing anything. Then classify the root cause, apply the smallest fix targeting only that cause, and rerun the original command to verify the check is green.

How to diagnose a flaky test versus a real code defect?▼

Reproduce the failure locally when it is deterministic and cheap, then classify it as code defect, test defect, flaky test, environment, dependency, or CI configuration. If the cause is unclear, gather more evidence from logs before changing any code.

Can this fix CI-only failures that cannot be reproduced locally?▼

Yes. When a failure is CI-only, the workflow inspects the CI log directly and records why local reproduction is not equivalent, then verifies the fix by citing the exact CI rerun or log.

When should I not use a CI-fixing workflow?▼

Do not use it for general feature work, broad cleanup after CI is already green, or guessing at failures without reading logs. Product-behavior changes beyond the failing check belong in a feature-development workflow, and broader coverage work belongs in a test-coverage workflow.

Why does the fix require a production-handoff record?▼

The handoff record documents changed files, base and head refs, summary, tests referenced and executed, residual risk, and required review lenses. This ensures the fix is reviewed by others rather than self-approved.