<![CDATA[The rapid expansion of global software infrastructure has created a critical bottleneck, as human developers struggle to manage escalating technical debt and complex repository maintenance. This research explores the transformative shift toward “Autonomous Repository Management” (ARM), where AI agents transition from passive assistants to independent maintainers. The primary objective is to evaluate the efficacy of agentic architectures in performing end-to-end maintenance tasks across diverse software ecosystems. Employing a longitudinal experimental design, this study utilized a purposive sample of 50 open-source repositories, applying a custom “RepoHealth-Bench” framework to measure performance. Findings indicate that AI agents reduced technical debt by 31.5% in legacy systems and achieved a 96.5% patch success rate in standardized libraries, significantly outperforming human-centric benchmarks in speed and security remediation. Inferential analysis reveals a strong correlation between repository documentation quality and agent reliability, suggesting a “compounding health” effect through iterative machine-led refactoring. The study concludes that the “Coding Revolution” effectively reverses software entropy, shifting the developer's role from manual execution to high-level orchestration. These results provide a foundational blueprint for integrating autonomous digital workforces into the modern software development lifecycle, marking the end of the manual maintenance era.]]>
Copyrights © 2025
