Skan is a Python library to analyze skeleton images, such as images of branching neurons, or of the molecular skeleton of a cell. It is written using NumPy, SciPy, and pandas, with key functions compiled by Numba for speed. I'll briefly describe the library and application, but then spend most of the time discussing Skan's gritty innards, including: how to write n-dimensional image analysis code instead of baking in 2D- or 3D-only logic; examples of using Numba to speed up real-world array-based code; why you should know about memory layout; and the versatility of SciPy's sparse matrix formats.