Anthropic Adds Plugin Evals to Claude Code: 6 Grader Types, a No-Plugin Baseline, and a CI Gate for Skills

Anthropic has introduced a new evaluation workflow for Claude Code that allows developers to test how specific plugins impact model performance. By comparing plugin-enabled outputs against a baseline, this tool helps determine whether additions improve or hinder the model's responses. The system includes six different grading types and a continuous integration gate to standardize the verification of custom capabilities.