run-tests

You can automate testing for Copilot Studio agents by running batch test suites or importing evaluation results to automatically identify failures and propose YAML configuration fixes.

To analyze Copilot Studio evaluation results, you can import CSV files from built-in evaluations to automatically identify issues like topic mismatches, response failures, or generative answer problems.

Yes, you need the Copilot Studio Kit installed and access to your agent's test configurations to run batch test suites, though you can also analyze CSV results from evaluations without it.

Copilot Studio agent tests typically fail due to topic mismatches, response issues, or generative answer problems, which can be identified through results analysis and fixed using proposed YAML changes.

Yes, automated fix proposals are generated as YAML configuration changes to directly address identified test failures and improve your Copilot Studio agent's performance.

The best way to identify topic mismatches is to analyze evaluation results or run batch test suites, which automatically detects failures and pinpoints the root cause of the agent's behavior.