Hey, let me start by saying that I really like cursor and I am using it every day for work now. However there is one issue that is making it quite frustrating sometimes, precisely commands run by the agents.
On the fisrt glance they are working, but there are issues like:
- sometimes agents run some stupid commands or they are making same mistakes over and over again (like trying to access a specific catalog on the disk already being in that directory) or they are completely unaware of the shell they are using (they are trying to use some shell specific pipelining methods) - it would be great if they were smarter about that or if I had more control over those commands (for instance I could edit them)
- There is no way (or I have not found one) to conveniently manage long running commands (like commands that are running local server) or stop them - it is causing issues like ports not being free and also resource usage is skyrocketing with every agent message. This is especially annoying when some models try to run commands like npm run dev every time they want to finish their response
Are there any tricks I can use right now to fix those issues? Or maybe there are some settings that I have not yet discovered? Either way it would be great to see those things fixed in the future.