Show HN: A pre-execution guard that stops AI agents running destructive commands

It looks useful; I am definitely going to try it.

Really useful!