Study Explores Reusable AI Agent Skills in Software Engineering

This research investigates the emerging trend of AI agents transforming traditional software engineering (SE) activities into reusable "skills." Historically, SE has advanced through increasing levels of reuse, from code libraries to services. With the rise of AI agents, the concept of packaging and sharing SE expertise as discrete, reusable skills is gaining traction. The study conducted a large-scale empirical analysis of SE skills found in public repositories and marketplaces. It examined the types of activities these skills encapsulate, their coverage across the software development lifecycle, how they evolve, and their evaluation methods. Key findings indicate that SE activities are indeed becoming more modular and reusable through agent skills. The research also highlights opportunities for developing better skill recommendation systems and more structured engineering approaches for these skills, particularly for complex, high-context SE tasks.