Adaptation of Task Goal States from Prior Knowledge