Docker is widely adopted for packaging and deploying applications in portable containers. Creating optimal Dockerfiles can be challenging due to varying best practices and deviations known as “Docker smells.” Existing tools for detecting Docker smells suffer from limited recognition and maintenance issues. In this study, we introduce a new tool, Parfum, that automatically detects and repairs Docker smells. We evaluate the effectiveness of Parfum and investigate the impact of Docker smells on image sizes in open-source Dockerfiles. Our results show that Docker smells affect image sizes. On average, eliminating Docker smells results in a size reduction of 46.38MB per image, leading to a total reduction of 39.93T in transferred data per week on DockerHub. We also identify packaging manager commands as the most impactful smells. We initiated 34 pull requests, 24/25 have been merged, and 9 pull requests waiting for an answer, indicating developers’ interest in repairing Docker smells.