Competence management in collaborative networks depends on the identification of gaps between the actual performance and the desired behavior. Therefore, collaborative networks need means to express their goals and intentions as well as to measure this against their own performance. In this paper, we present the i*-based modeling framework for agent-oriented elicitation of dependencies and goal-orientation in collaborative networks. Moreover, we sketch a method to translate the resulting models into a pattern-oriented analysis framework, which allows to compare the models with social network analysis results on real media traces left by the networks while collaborating using social software. This results in the identification of hidden or missing competences within a network. We will illustrate the approach with examples from ongoing project work. Finally, we discuss a collection of collaborative patterns indicating which (meta-)competences can be acquired.