evaluate-cortex-agent

What problem does it solve?

This Skill provides a structured, repeatable workflow to evaluate Cortex Agents using Snowflake’s native Agent Evaluations, enabling objective benchmarking and comparison of agent performance across configurations.

Core Features & Use Cases

• Define evaluation datasets for Cortex Agents and track metrics such as correctness, tool_selection_accuracy, tool_execution_accuracy, and logical_consistency.
• Automate setup of evaluation runs in Snowflake and generate Snowsight reports.
• Support scenario-based comparisons to measure improvements after prompts, tool changes, or configuration updates.

Quick Start

Configure the target agent, select metrics, build or choose a dataset, run the evaluation, and review results in Snowsight.