How do I convince Cursor to look at the terminal after running a command?

Support Help
1

I’ve created rules, memories and even tried yelling at it but it continues to run terminal commands that show errors but then it says everything worked (perfectly). I need to manually stop it and tell it to look at the terminal output despite the terminal having being spawned by Cursor.

Are there any tricks to getting this to stick?

2

Hi @MarcW usually there is no need to even mention to Agent that it needs to look at a terminal.

In last week I ran several parallel agents to create features from FRDs with extensive testing through CLI and haven’t had to tell an Agent more than 100% test coverage is required once it reached 99.8%.

Could you share more what you are trying to do and what goes wrong?

There are several likely causes but it really depends on your process.

  • Which model are you using?
  • Do you use rules? And if yes, how long or detailed?
  • Long chats seem to be a likely cause of this issue as agents eventually run into a lot of context with conflicting information.

I recommend stepping back for a second. Check what goes right and what goes wrong.

  • Keep chats short and focused on a single task.
  • Don’t attach too many files or rules, MCPs,… as it pollutes context.
3

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.