> nowhere near readteaming

I think any security-related task triggers it to think about the threat model and thus hit the guardrails