My point is that if the threat model is “agent got sensitive information and wants to exfiltrate it”, looking at the code isn’t going to save you. You’re alread

My point is that if the threat model is “agent got sensitive information and wants to exfiltrate it”, looking at the code isn’t going to save you. You’re already dead. This was the threat outlined in TFA.