The Agentic Attacker: One Objective, One Prompt, Forty Minutes, Domain Admin – Game Over

This research demonstrates in a controlled Active Directory lab that an agentic attack stack combining a frontier model (GPT-5.5), agent orchestration platforms, MCP-enabled tooling, and structured operational guidance can autonomously execute a complete attack chain from external access to Domain Administrator in as little as 40 minutes, highlighting that orchestration and tooling accelerate and automate existing attack workflows rather than solely novel exploit discovery.