The purpose of recommending activities to learners is to provide them with resources adapted to their needs, to facilitate the learning process. However, when teachers face a large number of students, it is difficult for them to recommend a personalized list of resources to each learner. In this paper, we are interested in the design of a system that automatically recommends resources to learners using their cognitive profile expressed in terms of competencies, but also according to a specific strategy defined by teachers. Our contributions relate to (1) a competency-based pedagogical strategy allowing to express the teacher’s expertise, and (2) a recommendation process based on this strategy. This process has been experimented and assessed with students learning Shell programming in a first-year computer science degree. The first results show that (i) the items selected by our system from the set of possible items were relevant according to the experts; (ii) our system provided recommendations in a reasonable time; (iii) the recommendations were consulted by the learners but lacked usability.