Background: Endoscopic surgery has been widely adopted in many surgical fields, increasing the importance of the education of surgical skills. Although simulation-based training is mainstream, training in the actual surgical environment is also indispensable. Objective: This study aims to develop an automated feedback system using kinematic data to enhance the efficiency of skill education for endoscopic sinus surgery (ESS). Methods: Kinematic data were collected for four basic scenes of navigation-guided ESS. Features of the surgical techniques were extracted from the collected data and compared with standard process models of experts to automatically identify technical issues. A system was constructed to classify these issues and provide feedback to surgeons using machine learning (random forest algorithm). Results: The developed system classified surgical issues with high accuracy and received high ratings in a survey conducted with surgeons. Conclusion: The proposed system is useful for post-operative review and can potentially enhance the efficiency of skill acquisition. It is expected to be applicable to other surgical procedures in the future.