Action languages allow to formally represent and reason about actions in a highly declarative manner. In recent work, revision and management of conflicts for domain descriptions in such languages wrt. semantic integrity constraints have been considered, in particular their reconciliation. However, merely ad hoc tests and methods have been presented to aid the user in analyzing and correcting a flawed description. We go beyond this and present a methodology on top of such tests for identifying a possible error, which works in several stages. The issue of such a methodology for action languages is novel and has not been addressed before, but is important for building tools and engineering action descriptions in practice.