Seems the agent has issues reading results that has been piped to cat or that results in a pager. Has anyone else experienced this? It just waits for the result indefinitely.
Instruct it not to use pagers etc, git --no-pager etc
I’ve been doing that for quite some time, but as we know, Cursor – or rather the LLM – has a tendency to forget instructions every now and then. Also this feels like a bandaid solution.
I think this is essentially an issue that should be resolved by the Cursor team, that is why I posted about it here. Not sure they have any other issue tracker I can use for this?
I mean - put it into “Rules for AI”. Helps alot. Also ensure your default terminal is not too “fancy” for AI consumption. On windows prefer powershell.
Long commands (many args) can be put into package.json scripts with descriptive name and AI can use it effectively via “shortcut”.