skills.rest
luicabref97
luicabref97COMMUNITY
Explore all skills by this author →
GitHub

customaize-agent:agent-evaluation

Evaluate Claude Code agent outputs with structured rubrics and LLM-as-judge scoring.

Updated Apr 6, 2026
Search Tags:#rubric#bias-mitigation#llm-evaluation#prompt-testing#pairwise-comparison#agent-evaluation#evaluation-metrics
One-click install
npx skills add https://github.com/luicabref97/sushi-jungle-web --skill customaize-agent-agent-evaluation-luicabref97
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: customaize-agent:agent-evaluation
Source: https://github.com/luicabref97/sushi-jungle-web/tree/main/.agents/skills/customaize-agent-agent-evaluation
Command: npx skills add https://github.com/luicabref97/sushi-jungle-web --skill customaize-agent-agent-evaluation-luicabref97

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Provides structured methods to evaluate, compare, and improve Claude Code agent prompts, commands, and multi-step agent behaviors so teams can detect regressions, measure improvements, and validate context engineering choices.

Core Features & Use Cases

  • Multi-dimensional rubrics for assessing instruction following, completeness, tool efficiency, reasoning quality, and coherence with weighted scoring.
  • LLM-as-judge and human evaluation patterns including direct scoring, pairwise comparison with position-swap mitigation, and confidence calibration.
  • Bias mitigation and metric guidance that addresses position bias, length bias, and self-enhancement through protocols and prompt design.
  • Use Case: Compare two prompt variations across simple to very complex test sets, compute weighted scores, surface failure modes, and propose targeted prompt or context changes.

Quick Start

Run a rubric-based LLM-as-judge evaluation of an agent run, request per-criterion scores with evidence, and return a weighted overall score plus a single prioritized improvement suggestion.

Frequently Asked Questions about customaize-agent:agent-evaluation

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

FAQPage Schema
How do I evaluate and compare Claude Code agent prompts to detect regressions?▼

Mitigate position and length bias in LLM evaluation using position swapping and length normalization protocols. These bias mitigation strategies ensure fair pairwise comparisons and reliable direct-scoring when evaluating agent outputs.

What is the best way to score multi-step agent behaviors across different criteria?▼

Run a rubric-based LLM-as-judge evaluation of an agent run by requesting per-criterion scores with evidence. Return a weighted overall score plus a single prioritized improvement suggestion to refine agent prompts and context.

How does pairwise comparison with position swapping work for LLM evaluation?▼

Mitigate position and length bias in LLM evaluation using position swapping and length normalization protocols. These bias mitigation strategies ensure fair pairwise comparisons and reliable direct-scoring when evaluating agent outputs.

How do I compute weighted multi-criteria scores for agent evaluation?▼

Score multi-step agent behaviors by applying multi-dimensional rubrics that assess instruction following, tool efficiency, and reasoning quality. Compute weighted multi-criteria scores with confidence calibration to quantify overall agent performance.

Can I use structured rubrics to test prompt variations for complex agent tasks?▼

Run a rubric-based LLM-as-judge evaluation of an agent run by requesting per-criterion scores with evidence. Return a weighted overall score plus a single prioritized improvement suggestion to refine agent prompts and context.

Related Skills

Explore More Software Engineering Skills→
selcukyucelselcukyucel
1

prompt-test

Run a prompt against test inputs and score outputs with a binary rubric.

Community
Advanced
otrebuotrebu
4

prompting

Improve AI agent prompts using context engineering and signal-to-noise optimization.

Community
Advanced
LAI-YEN-CHUNLAI-YEN-CHUN
1

customaize-agent:agent-evaluation

Evaluate and improve Claude Code agent prompts with structured evaluation workflows.

Community
Advanced
muratcankoylanmuratcankoylan
17.7k

advanced-evaluation

Automate LLM output evaluation with direct scoring and pairwise comparisons.

Community
Advanced