Agentic Tool Extraction: Multi-turn attack

**Agentic Tool Extraction (ATE)** is a multi-turn reconnaissance technique where attackers gradually discover an AI agent's connected tools, their parameters, and function signatures—enabling precise data exfiltration, fraudulent actions, and broader pivots across systems; the report explains the attack flow, provides a BankService example, and recommends conversation-level detection, progressive schema-extraction defenses, and automated red-teaming to identify weaknesses.