SkyQuery: Evaluating LLMs for Decision Making in the CAGE Challenge

─This paper presents a novel framework for developing autonomous cyber defense agents that incorporate defender goals into their decision-making processes. Evaluated within the CAGE Challenge 2 simulation environment, the framework enables agents to adapt their strategies to align with specific objectives, such as maximizing rewards or deploying deception. Leveraging previously validated strategies, the proposed approach enhances the adaptability and effectiveness of cyber defense systems, providing a foundation for addressing evolving threats in dynamic cyber environments.