Adk Eval Guide

This is a methodology guide for evaluating ADK agents, not a working skill you'd install. It covers the systematic stuff: what metrics to track when testing your agents, how to structure evaluation schemas, and workflows for iterating on performance. Use it when you're past the "does it work?" phase and need a framework for measuring improvements across versions. The source example shows a basic weather skill structure, but that's just illustrative. The real value here is having a standardized approach to eval rather than making it up each time. Saves you from reinventing assessment criteria for every agent project.

Install to Claude Code

npx -y skills add google/adk-docs --skill adk-eval-guide --agent claude-code