The use of a software reference architecture limits possible deviations and errors in the implementation of software projects, as the code must follow predefined rules that developers must respect to guarantee quality. However, when introducing new code to projects these rules can be violated. As a result, architectural erosion, bad smells, or even bugs that can be difficult to find are introduced to the projects. This paper proposes an approach for reviewing compliance to predefined rules that map architectural decisions to code. During the continuous integration process, the automatic analysis raises an issue for each rule violation. Developers can analyze and correct issues, and trace/visualize improvements, or lack thereof, through time. We present a validation experiment carried out in the context of a Software Development course, and we show how the approach helps developers to write better code.