threat modeling

Reciting all ten gets you a pass on memory and a fail on judgment. The question is really a prioritization exercise, here's the two-risk answer that maps to how agents actually get breached, and the trap hiding in 'walk me through.'

ATLAS questions filter candidates who can name the framework from those who can run an incident through it. The mapping exercise, one concrete attack, kill-chain stage by stage, is what the interviewer actually wants to hear.

When agents call agents, one agent's output is another's untrusted input, and a single injected instruction can propagate across the whole graph with accumulated privileges. The strong answer treats every inter-agent edge as a trust boundary and bounds privilege per agent, not per system.