Does cursor ask mode do anything better than agent mode? I feel like I just default to agent mode for everything; does anyone actively use cursor ask mode (when asking questions) and find it to be better than agent mode?
I used to use Ask mode extensively when I was having a discussion with the model about options, understanding code, or planning, etc.
This ensured that the model would not make changes to the code, and would only give advice.
The only thing that bothered me about Ask mode is that it takes a couple of clicks (on the tiny UI elements) every time I needed to switch between Agent/Ask.
More annoyingly, when I was ready to implement a solution, I would often forget to switch from Ask to Agent which meant that it would run through the complete solution, but not update the code, so I would then have to switch to agent mode and run the request again.
When Commands were implement in Cursor, I stopped using Ask entirely - I now have a command:
/adviceonly
… which looks something like this:
”Do not update any code - just provide information based on the prompt.”
I was concerned that the model might sometimes ignore the request to not update code, but I have found it 100% reliable.
Now I just have to type /a[Enter] (to select the /adviceonly command) at the end of any prompt, and it acts like it is in Ask mode.
I only have 6 commands, but I use them extensively. Examples are:
/buildcheck
/adviceonly
/summarise (this command summarises my prompt to ensure that the agent has understood the request, and also presents options and also prompts for clarifications - I use this very frequently)
1 Like
When I want to for sure know it won’t edit any code. I often asking it to analyze something, propose changes, or just using it like chatgpt/google.