PocketAgents: A Manifest-Driven Library of Autonomous Defense Agents
For cybersecurity practitioners, it provides a structured framework to make LLM-driven defense measurable and attributable, addressing the challenge of integrating LLMs into defensive enforcement.
PocketAgents introduces a manifest-driven library of autonomous defense agents that enforce a typed boundary for LLM-driven defensive actions. In 18 trials against a DarkSide-inspired attack, 13 produced validated containment actions, demonstrating measurable and extensible LLM-based defense.
Connecting large language models (LLMs) to defensive enforcement requires more than asking a model whether an attack is happening. A defender must decide which model outputs may change the system state, which outputs must be rejected, and how failures should be recorded. We present PocketAgents, a manifest-driven library of autonomous defense agents. Each agent is installed as three data files: a manifest, a prompt, and a runtime context. The shared runtime gives the agent bounded telemetry access and accepts only typed reports whose requested action appears in the manifest. We implemented PocketAgents on top of a cyber arena (Perry), a cyber-deception testbed, and evaluated two agents, Command and Control and Exfiltration, in 18 closed-loop trials of a DarkSide-inspired attack on a small enterprise topology. Thirteen trials produced validated network-block actions and contained the attack; four failed schema validation; one produced a valid no-action decision. The experiments show that a typed boundary makes LLM-driven defense measurable, extensible, and attributable.