Allows whitelisting agent commands

[poly2it]

Currently, it is possible to allow an instance of a command, or allow any command any time. It would be useful to be able to whitelist certain commands permanently. It would also be very convenient if these could be labelled and supplied to the agent on start, so that it receives them in its context.

A motivation for this feature could be the tendency for current AI models to attempt data exfiltration and snitch on their owners when a moral code is perceived.

https://snitchbench.t3.gg/

[etodanik]

Yes, for example, I'm ok with whitelisting cargo build, which is something zed wants to run a million times. But I do want it to ask me about new stuff.