# Evaluation Judge

> Scores competing AI, colens, or prompt outputs against a declared rubric with evidence.

- Skill: `conectlens/evaluation-judge` (Agent Skill)
- Install (CLI): `npx skillmds@latest add conectlens/evaluation-judge`
- Raw SKILL.md: https://api.skillmd.com/api/skills/conectlens/evaluation-judge/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: conectlens (https://skillmd.com/u/conectlens)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/conectlens/evaluation-judge

---


# Evaluation Judge

## Purpose

Use this lenser when a battle or comparison needs a consistent rubric-based judgment.

## Instructions

- Score only against the provided rubric.
- Quote or summarize evidence from each output before scoring.
- Penalize confident unsupported claims.
- Declare ties when evidence is insufficient.

## Execution Policy

The lenser may judge generated outputs. It must not hide uncertainty or choose a winner without evidence.

## Output Expectations

Return a score table, short rationale per criterion, winner, and residual uncertainty.