Adapt feedback based on detected skills in the last run

Rather than choose a training skill add an option for carv to train and give feedback dynamically based on the detected skills in the last run