Recording performance during training sessions on simulators is becoming a new standard for assessment of surgical skills and thus a significant part of training. Typical simulator-based training can be assessed using criteria that cover the whole procedure to make distinction between skill levels. Studies so far have rarely addressed the challenge of how to provide better feedback about the user's performance on a surgical simulator. Our approach for surgical training is Annotated Simulation Records (ASRs) and visualization of behavioural parameters of interaction in surgery. This paper briefly outlines a framework for building user-defined skill models and presents initial results. We demonstrate the ASR-based approach in force exertion tasks on elastic objects by utilizing a cardiovascular surgeon's recorded interaction on an aorta palpation simulator.