Even though human activities may involve physical injuries, there is not much discussion in the academy of how ubiquitous computing could assess the risk related to them. This paper proposes an approach to evaluate the risk of activities considering two factors: actions that compose activities and user performance in such activities. Risk management based on composed actions is measured through the analysis of the frequency of each action for a particular user, so that we are able to capture his customary behaviour. The evaluation of the user's performance is accomplished by addressing performance properties, such as: attention, duration, effectiveness, etc.Our work has its foundations in the Activity Theory for modeling activities allowing the mediation by tools of the interactions between the subject and the environment and in the behavioural model Skill-Rule-Knowledge for understanding the subject's behaviours. To validate our model we developed a case study to demonstrate the functioning of our work. In this scenario we analyze how activities are detected, actions are evaluated and the performance is inferred. At last, an analysis of the final risk for the activity is presented.