_agent-evaluation-performance

Evaluating agent definitions involves assessing length, density, and redundancy to prune unnecessary context and prevent duplication with loaded skills. This ensures every line earns its place and maintains efficient loading.

Context efficiency in agent definitions means minimizing unnecessary content and preventing duplication with loaded skills. It matters because lean definitions reduce context bloat, ensure appropriate model choices, and minimize tool grants and risk.

Auditing an existing agent definition requires providing the definition to receive a structured critique. The evaluation assesses skill loading efficiency, model appropriateness, and tool grants, then suggests concrete remediation steps to prune redundant guidance.

Yes, you can use this evaluation when reviewing a new agent to ensure it loads only necessary skills and avoids duplicating what the skill already provides. It exposes findings in a structured format with actionable remediation suggestions.

The best way to assess model appropriateness and minimal tool grants is to evaluate the agent definition across skill loading efficiency and tool permissions. This minimizes context usage and risk while aligning definitions with loaded skills.

You need to evaluate agent definitions for context efficiency when reviewing new agents to ensure proper skill loading, or when auditing existing agents to prune redundant guidance and align definitions with loaded skills to prevent context bloat.