Implementation of AI Agent remote terminal access

I have noticed that the AI sandbox prevents the AI agent from accessing any kind of remote ssh functionality, this is very restrictive when operating on remote interpreters. The only workaround seems to be storing an ssh-config in the project directory (unsafe). 

I feel it would be prudent to supply a tool that allows the developer to grant (single-command, or session-long) access to an internal ssh-config (settings >> tools >> ssh configurations). This would prevent any kind of credential leakage whilst allowing for full “human in the loop” control of the terminal activity.

Cheers.