The Shrinkwrap operation computes the convex hull of the input geometry using the same method as Tetrahedralize, and translates the resultant polygonal primitives inwards along their normals. Unlike the Peak SOP, which takes the point and moves it according to the average of the neighboring polygons, Shrinkwrap moves the polygons and recomputes their intersections, possibly creating new and removing old points, but maintaining polygon normal directions. Polygons with more than three edges are divided into triangles with the same normals as the original polygon.
The translation distance inwards.
The error tolerance for computing the new point positions.
The error tolerance for fusing points together. This is equivalent to using consolidate points in the Facet SOP.