A two-phase evaluation pipeline checks the work produced in this conversation. First, a meta-judge creates tailored scoring criteria. Then a judge applies those criteria with isolated context and evidence-based feedback. The result is a structured report only, with no automatic changes.
This helps you get clear, unbiased assessments of your code, documentation, or other outputs. Every score includes specific citations from your work, like file locations and line numbers. The evaluation focuses on what matters most for your specific task.
You can use it to review completed work or check specific aspects by providing an evaluation focus. The process prevents confusion from earlier conversation history, giving you a fresh and fair review.
Add Judge skill to your workflow
Global
mkdir -p ~/.claude/skills/judgeProject
mkdir -p .claude/skills/judgeSource Repository
More Code Review & Quality Claude Code Skills
Improve Codebase Architecturemattpocock/skills
Scan your codebase for architecture improvements with visual reports and pick one to refactor
Web Design Guidelinesvercel-labs/agent-skills
Review your UI code for compliance with web interface guidelines
Caveman Reviewjuliusbrussee/caveman
Ultra-compressed one-line code review comments that cut noise
Zoom Outmattpocock/skills
Get a big picture view of your code with a simple map of modules and callers
Codebase Designmattpocock/skills
Create deep modules with tiny interfaces and lots of hidden logic for testability
Requesting Code Reviewobra/superpowers
Catch code issues early with focused automated reviews before every merge
Request Refactor Planmattpocock/skills
Plan safe code refactors with tiny commits and clear GitHub issues
Receiving Code Reviewobra/superpowers
Think first, then code. Verify feedback before making changes. Stay objective