Skeleton is a set of Cocoa libraries optimized for fast development of research, and for professional applications dealing with the analysis of musical signals. Based on fundamentals of psychoacoustics, perception, and learning, the underlying framework consists of machine-listening and machine-learning tools, supported by flexible data structures and visualizations. Designed essentially as an alternative to more generic and slower tools such as Matlab, Skeleton should provide a robust, efficient, specific and yet open programming environment for deconstructing, structuring, and labeling audio, or for generating personalized music. Skeleton aims to combine various solid scientific approaches to music listening, learning, and cognition, to enable the development of consistent creative audio applications, and enhance music-making. Current applications include beat matching, music mosaicing, compression, and music textures.