I had Cursor Agent write some Python programs to clean data.
For a while, the agent worked well, but after a while, it always responded before the command gave any information, debug, or error information, so it couldn’t get the correct error messages.
If I accept the next command which the agent suggest in this instant response, it’ll response to the output the previous command gave.