Why is your agent able to call ssh. Why can it trigger 1password. Why are you giving metaphorical guns to metaphorical toddlers. Why is it not sandboxed. Your practices worry me.
Are you guys all running agents in VMs?
Are you guys all running agents in VMs?