Large language models (LLMs) pose unprecedented challenges to courses centered on experimental teaching in areas such as computer networks. Students may use LLMs to complete coding tasks and technical reports, which not only diminishes the effectiveness of experimental training but also impacts the fairness of assessments. To address this challenge, this paper proposes a “Three-Dimensional Competency Development” framework tailored for computer network experimental teaching. The framework reconstructs computer network experiments from three dimensions: consolidating foundational skills, enhancing analytical abilities, and elevating innovative capacities, through a progressive design. Implementing this experimental framework in specific teaching practices shows that over 70% of students can effectively grasp the underlying principles and knowledge of computer networks.