adhoc-testing-for-cortex-agent

Yes, the skill accepts optional experimental flags for staged reasoning and date overrides, allowing you to explore specific agent behaviors and debug edge cases during your interactive testing sessions.

The skill requires Python scripts and depends on the requests, snowflake-connector, and urllib3 libraries to function. You need an environment configured with these dependencies to execute the interactive testing sessions and connect to your data sources.

The best way to debug unexpected Cortex Agent outputs is to run targeted questions through the interactive testing script, review the responses directly, and use experimental flags to isolate reasoning steps or override dates for specific test scenarios.

The skill requires Snowflake connectivity, as indicated by its snowflake-connector dependency. You need an environment configured with this library to execute the interactive testing sessions and connect to your data sources for Cortex Agent validation.

Use adhoc testing for quick debugging, exploring agent behavior, and iterating on workflow logic. Switch to formal dataset curation when you need to persist test outputs into structured evaluation datasets for comprehensive regression testing.

Your testing workflow may not capture results because this skill is designed for interactive debugging rather than dataset persistence. To properly save generated test outputs, integrate with the dataset-curation skill for formal evaluation tracking.