False positive on detection of directories to be added to allow list

[Tommertom]

Hi all

I am coding a Telegram chatbot. As part of the workflow I am building commands that are trigger by typing a /. Example /help or /dothis.

Interestingly, when I ask copilot to add a new command (with or without the slash) then after a few agent iterations it wants to add the command to the allowed directory list. Which is wrong and undesirable as it may expose directories to the agent which I dont want (and can deny accordingly ofc)

Example prompt "Add command dothis to the bot which should do this and that when the user issues the command". Then later on it asks "do you want to add /dothis" to the allowed directories

I reckon the directory detector in the cli may need some adjustments in the (near) future

Thx for the great tool and keep up the good work!

[thaapaniemi]

I have encoutered the same problem when working with http API endpoints. Like access to fs root with //, or some endpoint path like /foo/bar.