Modern music streaming services are heavily based on recommen- dation engines to serve continuous content to users. Sequential recommendation—continuously providing new items within a sin- gle session in a contextually coherent manner—has been an emerg- ing topic in current literature. User feedback—a positive or negative response to the item presented—is used to drive content recom- mendations by learning user preferences. We extend this idea to the session-based recommendation domain to improve learning of context-coherent music recommendations by modelling negative user feedback, i.e., skips, in the loss function. To this end, we propose a sequence-aware contrastive sub-task to structure item embeddings in session-based music recommen- dation, such that true next-positive items (ignoring skipped items) are structured closer in the embedding space, while skipped tracks are structured farther away from all items in the session. Since this causes skipped item embeddings in a session to be farther than unskipped items in the learned space, this directly affects item rankings using a K-nearest-neighbors search for next-item recom- mendations, while also promoting the rank of the true next item. Experiments incorporating this task into SoTA methods for sequen- tial item recommendation show consistent performance gains in terms of next-item hit rate, item ranking, and skip down-ranking on three music recommendation datasets, strongly benefiting from increasing presence of user feedback.