We present a robust method for automatically constructing an ellipsoidal skeleton (e-skeleton) from an unorganized set of 3D points. This skeleton will be essentially useful for dynamic visualization, manipulation and deformation. It also provides a good initial guess for surface reconstruction algorithms. On output of the entire process, we obtain an analytical description of a solid, semantically zoomable (local features only or reconstructed surfaces), with any level of detail (LOD) by discretization step control in voxel or polygon format. This capability allows us to handle objects at interactive framerates once the e-skeleton is computed. To ensure robustness and accuracy, all points sampled from a solid are taken into account, including the inner ones. Each e-skeleton is stored as a multi-scale CSG implicit tree. We propose a data structure to store along not only extracted features and geometry but also the hierarchy of shape refinement. Applications cover a wide range in computer graphics, from CAD to medical imaging.