llm-judge

Name: llm-judge
Availability: InStock
Author: javierhbr

Compare code implementations across repositories using rubric-based scoring.

Updated Feb 17, 2026

Search Tags:#code quality #testing #code review #security analysis #llm-as-judge #repository comparison

Ready to run

npx skills add https://github.com/javierhbr/random-poc --skill llm-judge

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: llm-judge
Source: https://github.com/javierhbr/random-poc/tree/main/custom-skills/beagle-main/plugins/beagle-analysis/skills/llm-judge
Command: npx skills add https://github.com/javierhbr/random-poc --skill llm-judge

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @beagle:llm-artifacts-detection, and includes references (resource) components.

What problem does it solve?

This Skill automates the complex and time-consuming process of evaluating and comparing code implementations across multiple repositories based on predefined criteria.

Core Features & Use Cases

Multi-Repo Code Evaluation: Compares code quality, functionality, security, and more across different codebases.
LLM-as-Judge Methodology: Leverages AI agents to perform detailed, rubric-based scoring.
Structured Reporting: Generates a ranked report with detailed justifications for each score.
Use Case: When deciding which of several competing feature implementations to merge, use this Skill to objectively score each one on functionality, security, and maintainability.

Quick Start

Use the llm-judge skill to compare the code in '/path/to/repo-a' and '/path/to/repo-b' against the spec in '/path/to/spec.md'.

Frequently Asked Questions about llm-judge

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

FAQPage Schema

How do I use AI to compare code quality across multiple repositories?▼

AI-powered code comparison scores repositories by applying an LLM-as-judge methodology with weighted rubrics, evaluating functionality, security, test quality, overengineering, and dead code to generate a ranked report with detailed justifications.

What is the LLM-as-judge methodology for code review?▼

The LLM-as-judge methodology for code review applies structured fact-gathering and weighted rubrics to AI agents, scoring implementations across functionality, security, and maintainability to objectively rank competing feature implementations.

How do I objectively evaluate and rank competing feature implementations?▼

To objectively evaluate and rank competing feature implementations, you can use an AI-powered skill to compare code in specified repository paths against a spec file, generating a structured report with detailed score justifications.

Does llm-judge work with specific spec files for repository comparison?▼

Yes, repository comparison supports spec files by accepting a path to a spec markdown file alongside multiple repository paths, using that specification as the baseline for structured fact-gathering and weighted rubric scoring.

What criteria are used to score code implementations in automated code review?▼

Automated code review scores implementations using weighted rubrics across five criteria: functionality, security, test quality, overengineering, and dead code, providing detailed justifications for each score in a ranked report.