I am having a pretty consistent issue that when Cursor runs test scripts in Agent mode (using Claude 3.7 Sonnet), it will execute the script, the script will get back to a complete message and terminal prompt, and it will just hang there, I have to come back and click cancel. It does it near 100% of the time with this script. This is a pretty annoying bug as it stops the agent in its tracks and I have to manually move it forward (it will literally sit there all night until I hit cancel).
I have even added custom rules in Cursor Rules to try to get around it:
“* Script Completion Signaling:
* All standalone scripts (e.g., ingestion, data processing, specific tests run directly via python scripts/some_script.py
) MUST print a clear, standardized success message to standard output upon successful completion. Example: INFO - [SCRIPT NAME] - Script completed successfully.
Once any “SCRIPT COMPLETED” message is detected OR you see the command prompt again (ie /app#), complete/cancel the script and move forward.”
But it doesn’t seem to make a difference.
Windows, 0.48.8