CyberAGENTS Framework Enhances Gamified Cybersecurity Learning.

Gamified learning is highly effective for subjects like cybersecurity that demand active problem-solving and iterative skill development. While generative AI agents offer a scalable way to deliver adaptive learning experiences, they pose risks such as inconsistent behavior, AI hallucinations, and misalignment with established pedagogical frameworks. This research introduces CyberAGENTS to mitigate these issues. CyberAGENTS is an agentic framework designed for gamified cybersecurity learning, ensuring structured autonomy. It achieves this through three core mechanisms: ontology-guided validation, which enforces domain-consistent reasoning and safety; schema-governed behavioral control, which bounds agent autonomy while retaining generative flexibility; and competency-based progression, which structures learning topics by difficulty and prerequisites, aligning with scaffolded instruction principles. The learning process within CyberAGENTS is broken down into four specialized agents—challenge, support, evaluation, and reward—each operating under behavioral schemas that dictate their modes and progression logic. Classroom deployments with undergraduate students and expert evaluations confirm that these structured controls significantly improve learner engagement, feedback clarity, and trust in AI-generated responses, demonstrating a blueprint for pedagogically sound agentic learning systems.