Find and fix weak spots in your AI agent.
What it does
Evaluates AI models and agents on Agent Platform, identifies failures, and verifies improvements through repeated scoring.
When to use it
Use when creating evaluation datasets, evaluating models or agents, choosing metrics, analyzing failures, or checking improvements.
How to use it
Give it an evaluation request and the cloud project and location; it runs approved evaluations and saves JSON and HTML reports.
What you provide
- a Google Cloud project
- a Google Cloud location
Uses
AP
Good to know
- cannot be undone: Agent Platform
Latest review
What you need · 6
Requires Python to run the SDK and bundled evaluation scripts.
Requires the evaluation SDK package at version 1.163.0 or later; setup installs it if the import probe fails.
Access · 5
This skill
Agent Platform
Write
Sends data for paid evaluations (irreversible)
GCS
Read
Reads evaluation datasets from GCS