agent-evaluation

Evaluating AI agent and skill quality involves applying a standardized 6-step rubric to score structure, content, and integration gaps. This process audits items against a 100-point scale to surface missing YAML front matter, Operator Context, or error handling.

A rubric-based audit is a standardized evaluation process that checks Claude Code agents and skills for completeness and standards compliance. It applies a 6-step rubric to identify structural and content gaps, producing traceable results with file paths and line references.

Auditing a collection of agents for missing error handling requires running a standardized evaluation workflow. The rubric checks for anti-patterns, validates references, and outputs a structured report with prioritized findings and recommended improvements.

Yes, you can score individual agents against a 100-point quality standard by applying the evaluation rubric. The audit objectively scores the agent's YAML front matter, Operator Context, and error handling to identify specific quality gaps.

A structured agent evaluation report includes prioritized findings, recommended improvements, and traceable results with specific file paths and line references. It documents quality gaps found across the collection or individual items during the rubric assessment.

You need to run an agent quality assessment when auditing newly created or updated agents and skills, or when assessing the overall health of your collection. The evaluation ensures your agents meet structural and content standards before deployment.