Automatic assessment has recently drawn the efforts of researchers in a number of fields. While most available approaches deal with the construction of question items that assess factual and conceptual knowledge, this paper presents a method and a tool for generating questions assessing procedural knowledge, in the form of simple proof problems in the domain of the Euclidean Geometry. The method is based on rules defined as Horn clauses. The method enumerates candidate problems and certain techniques are proposed for selecting interesting problems. With certain adaptations, the method is possible to be applied in other knowledge domains as well.