Generating Context-Aware Learning Materials for Software Security via LLM Agents and Traceability

This paper presents a prototype that embeds security-focused, contextualised learning directly into the software-development workflow. By using trace links which already bind security standards, risk analyses, and vulnerability scan results. The prototype takes advantage of a multi-agent AI architecture to turn a security problem description and trace links from real life cases from a company into tailored learning material, lesson plans, and multimedia examples. This design positions real-life evidence at the centre of instruction and shows how LLMs can scale high-quality security training across projects. Questionnaire studies with professional developers reveal the educational impact of the LLM-generated learning material. Participants improved significantly between pre- and post-report assessments, with statistical tests confirming the gains. Furthermore, the learning material was subjectively rated above expert-written equivalents for clarity and relevance by questionnaire participants. Perceived quality correlated with actual learning, underscoring the pedagogical soundness of the contextualised, LLM generated content. Finally, we discuss the potential for integrating such adaptive, trace-link powered training examples into everyday development practices and suggest a portable framework for broader industrial adoption.