agent-evaluation

Evaluate Claude Code commands, skills, and agents with structured rubrics.

7|3|Updated Jan 15, 2026

Search Tags:#metrics #claude-code #prompt-engineering #bias-mitigation #evaluate #llm-judge

Ready to run

npx skills add https://github.com/Zpankz/mcp-skillset --skill agent-evaluation-zpankz

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agent-evaluation
Source: https://github.com/Zpankz/mcp-skillset/tree/main/agent-evaluation
Command: npx skills add https://github.com/Zpankz/mcp-skillset --skill agent-evaluation-zpankz

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Evaluate and improve Claude Code commands, skills, and agents. It helps teams test prompt effectiveness, validate context engineering choices, and measure improvement quality.

Core Features & Use Cases

Evaluation rubric design and application across single and multi-agent prompts.
LLM-as-Judge patterns and bias mitigation techniques for robust assessments.
Structured reporting, trend analysis, and guidance for prompt refinement.
Reusable test sets and hierarchical evaluation workflows.

Quick Start

Run an evaluation pass on your latest Claude Code prompt to identify gaps, surface actionable improvements, and document the results.

Frequently Asked Questions about agent-evaluation

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

FAQPage Schema

How do I evaluate and improve Claude Code agents using LLM-as-Judge patterns?▼

To evaluate Claude Code agents, define evaluation rubrics and apply LLM-as-Judge patterns with bias mitigation techniques. This delivers structured outputs and provenance for iterative prompt refinement and performance measurement.

What is an evaluation rubric for prompt engineering and how does it work?▼

An evaluation rubric for prompt engineering defines specific criteria to assess agent performance and context validation. It works by applying these metrics across single or multi-agent setups to measure improvement quality and surface actionable gaps.

Can I use this approach to validate context engineering choices for multi-agent setups?▼

Yes, you can validate context engineering choices for multi-agent setups by applying hierarchical evaluation workflows. It measures improvement quality across multiple agents and delivers structured reporting for prompt refinement.

What's the best way to mitigate bias when using LLM-as-Judge for agent evaluation?▼

The best way to mitigate bias in LLM-as-Judge evaluations is to apply structured bias mitigation techniques alongside defined rubrics. This ensures robust assessments and delivers provenance for iterative prompt improvement.

How do I document evaluation results and track prompt improvements over time?▼

Document evaluation results and track improvements by generating structured reporting and trend analysis. These outputs provide provenance and actionable guidance for iterative prompt refinement across test sets.