google-agents-cli-eval

To evaluate AI agent performance, you can use built-in and custom evaluation metrics to measure behavior across various dimensions. This Skill provides a suite of Python-based tools to generate traces and analyze your agent's performance on a given dataset.

Yes, you can generate synthetic datasets to test your LLM agent. The data synthesis feature creates synthetic user scenarios, allowing you to test agents in different conversational contexts before deployment.

The best way to optimize agent prompts is by using dedicated optimization tools that enhance performance based on specific metrics. This allows for iterative agent development and targeted performance analysis after running initial evaluations.

Yes, you need the google-agents-cli dependency installed to execute agent evaluations and optimizations. This Skill requires the agents-cli environment to generate traces and apply performance metrics to your AI agents.

When analyzing agent performance with synthetic data, limitations may include scenarios that do not fully capture real-world conversational complexity. Use the optimization tools iteratively to refine metrics and ensure the agent meets actual quality standards.