Verified:GitHub

ci-diagnostician

Diagnose CI build failures and propose minimal fixes from GitHub Actions and GitLab logs.

4|2|Updated Jan 27, 2026

Search Tags:#root cause #remediation #github actions #diagnosis #gitlab ci #build failure #ci

One-click install

npx skills add https://github.com/Arete-Consortium/ai-skills --skill ci-diagnostician-arete-consortium

Or copy as Structured Prompt for Agent▼

Please help me install this Agent Skill.
Skill: ci-diagnostician
Source: https://github.com/Arete-Consortium/ai-skills/tree/main/agents/analysis/ci-diagnostician
Command: npx skills add https://github.com/Arete-Consortium/ai-skills --skill ci-diagnostician-arete-consortium

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Red CI builds force engineers to dig through noisy logs to find root causes. This Skill fetches failed CI logs, extracts failure signatures, classifies them with a structured taxonomy, and produces an evidence-based diagnosis with a minimal, human-approved fix. ## Core Features & Use Cases - Log Ingestion & Parsing: Downloads and normalizes failed job logs from GitHub Actions, GitLab CI, or local files, then extracts structured failure signatures with file and line references. - Taxonomy-Based Classification: Tags each failure with technical, content, and CI-specific modes (test_regression, dependency_failure, infra_timeout, secret_leak, and more) with justification. - Guided Remediation: Proposes minimal patches only after explicit human approval, verifies them by running local tests, and escalates security-related failures immediately. - Use Case: A Dependabot PR bumps pydantic to v2 and the auth test suite goes red. The Skill fetches the failed run's logs, identifies the broken BaseModel import, classifies it as a dependency_failure, and proposes a two-line compatibility patch after your approval. ## Quick Start Ask the agent to diagnose the failed CI run for your repository by providing the owner, repo name, and run ID, and request a root-cause diagnosis with a recommended fix.

Frequently Asked Questions about ci-diagnostician

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

FAQPage Schema

How do I diagnose a failed GitHub Actions workflow?▼

Provide the repository owner, repo name, and failed run ID. The Skill downloads the failed job logs, extracts failure signatures with file and line references, classifies the failure mode, and returns a diagnosis with severity, likely culprit commit, and a recommended fix.

How to find the root cause of a flaky CI test?▼

The Skill parses logs for timeout and intermittent failure patterns, then classifies them as flaky or infra_timeout modes. It distinguishes genuine code regressions from infrastructure noise using log evidence rather than blaming CI providers without proof.

Does it support GitLab CI or only GitHub Actions?▼

It supports GitHub Actions, GitLab CI, and local build logs. Set the log_source input to github, gitlab, or file depending on where the failed run's logs come from.

Can it automatically fix CI failures without approval?▼

No. Every patch requires explicit human approval before being proposed or applied. Fixes are verified by running the relevant local test suite, and any regression causes an automatic revert and abort.

When should I not use automated CI failure diagnosis?▼

Skip it when the failure is obvious from a quick read, when the fix requires large architectural changes, or when no build logs are available. Without logs, diagnosis degrades to speculation.

What happens when a CI failure involves a leaked secret?▼

Security-related failures such as secret leaks or unauthorized API calls stop all diagnosis work immediately and are escalated to the user. The Skill never attempts to patch security failures autonomously.