When I ask questions using the GPT or Claude models of the agent, it can provide analysis and feedback based on my questions and the modifications have been completed. However, the code itself has not been actually changed.