evaluation

Evaluate agent performance using multi-dimensional rubrics and weighted scores.

Updated Jan 8, 2026
One-click install
npx skills add https://github.com/yeeehaooo/agent-kit --skill evaluation-yeeehaooo
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: evaluation
Source: https://github.com/yeeehaooo/agent-kit/tree/main/.agent-skills/skills/evaluation
Command: npx skills add https://github.com/yeeehaooo/agent-kit --skill evaluation-yeeehaooo

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This skill provides a structured framework to evaluate agent performance, enabling objective, reproducible quality assessments across iterations and deployments.

Core Features & Use Cases

  • Multi-dimensional rubrics that assess factual accuracy, completeness, citation quality, source credibility, and tool efficiency.
  • Evaluation pipelines that combine automated checks with human review, test set management, and production monitoring to track performance over time.
  • Use cases include validating context strategies, benchmarking agent configurations, and enforcing quality gates in continuous deployment.

Quick Start

Install or run the evaluation framework and load a prepared test set to begin scoring agent outputs against defined rubrics.

Frequently Asked Questions about evaluation

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

FAQPage Schema
How do I evaluate agent performance using rubrics?

Agent performance evaluation uses a deterministic rubric-driven evaluator and test-set management to score outputs against multi-dimensional rubrics. It assesses factual accuracy, completeness, citation quality, source credibility, and tool efficiency to compute weighted scores from per-dimension results.

What is multi-dimensional rubric evaluation for AI agents?

Multi-dimensional rubric evaluation assesses agent outputs across factual accuracy, completeness, citation quality, source credibility, and tool efficiency. It combines automated checks with human review to provide objective, reproducible quality assessments across iterations and deployments.

Can I set up production monitoring to track agent quality over time?

Yes, production monitoring is supported through built-in monitoring hooks within the evaluation pipeline. These hooks track agent quality over time by combining automated checks with test set management to monitor performance across dashboards.

How do I validate context engineering strategies for agents?

Validate context engineering strategies by running the evaluation framework against a prepared test set. This benchmarks agent configurations by computing weighted scores from per-dimension results using the deterministic rubric-driven evaluator.

Does this evaluation framework enforce quality gates for continuous deployment?

Yes, the evaluation framework enforces quality gates in continuous deployment by combining automated checks with human review. It uses test set management and production monitoring to ensure objective, reproducible quality assessments before deployment.

What are the limitations of rubric-driven evaluation for agents?

Rubric-driven evaluation relies on predefined dimensions like factual accuracy and tool efficiency, meaning unanticipated agent behaviors outside the rubric scope may go unscored. It requires prepared test sets and combines automated checks with human review to mitigate gaps.