AML.T0084

MITRE ATLAS / Discover AI Agent Configuration

Adversaries locate and read the configuration of AI agents on a victim system, learning which tools each agent may call, what its instructions say, and how much autonomy it has.

Think of it likeReading a new colleague's job description to learn precisely what they are authorized to do.

In plain English

Agent config is a map of what is reachable. Knowing an agent's tools and permissions tells the attacker exactly what they gain by taking control of it.