Text this: An empirical study of testing practices in open source AI agent frameworks and agentic applications