Agent Output Evaluation & Scoring

This skill implements a rigorous LLM-as-judge framework designed to verify the quality of agent outputs before task completion. By scoring content across five critical dimensions—Accuracy, Groundedness, Coherence, Completeness, and Helpfulness—it provides a structured, data-driven approach to AI validation. It replaces subjective assessments with nuanced weighted composite scores and mandatory evidence citations, ensuring that every claim made by an AI agent is verifiable, internally consistent, and actionable within your specific codebase context.

主要功能

使用场景