google-agents-cli-eval

Find agent failures and prove each fix.

agent evaluation

What it does

Evaluates agents, diagnoses failures, and iterates code or prompt fixes while comparing results.

When to use it

Use when evaluating an agent, creating evaluation data, analyzing failures, comparing results, or optimizing agent quality.

Give it an agents-cli project; it prepares evaluation data, runs and grades evaluations, diagnoses failures, and applies verified fixes.

What you provide

This skill

Vertex Evaluation Service

Write

Sends traces for cloud grading

Requires agents-cli; the documented installation command is uv tool install google-agents-cli.