nx_skinwrap.nx

nx_skinwrap.nx -- SKIN AS AN ENVELOPE OVER BONE, offset by FORENSIC SOFT-TISSUE DEPTH. ★WHY: pipeline stage 4. This program's skin has always been emitted INDEPENDENTLY of the layers under it, which is why brow/cheek/jaw/chin had to be modelled as parts STUCK ON an ovoid and why three tuned chin attempts all rendered as stacked lobes. The operator named the root cause -- "you aren't building from the inside out" -- and the fix is not another feature part: it is to DERIVE the surface from the bone beneath. With a generated skull now sitting in the body's bone layer, that is finally possible. ★THE RULE IS MEASURED, NOT TASTED. Forensic facial reconstruction (Rhine / De Greef class) publishes the millimetres of soft tissue over bone AT EACH LANDMARK. This codebase already owns that table -- it was ported once into knowledge/face_tissue.dat and its first two values alone (brow at the glabella, eye line) moved the binding number 92 -> 138 without a line of new geometry. Those depths are ABSOLUTE physical measurements, which is exactly the kind of value that TRANSFERS between heads where a y-position does not. ★AUDIT RAN FIRST, and it changed nothing but is why this is a new organ rather than an edit: nx_meshthick MEASURES wall thickness by ray-casting and is built on nx_mesh3, which is capped at 4096 verts and cannot hold a 43k-triangle skull. nx_mesh_sculpt has an inflate brush that pushes verts along their normal, but it is RADIUS-LIMITED around a brush point on a different mesh kernel. Neither does a global, per-landmark-depth offset on NXMSH2. The "push along the normal" idiom is borrowed from nx_mesh_sculpt deliberately, so this is consistent with the ecosystem rather than a second dialect. nx_skinwrap <out.nxmesh> <bone.nxmesh> [depth_scale_permil] nx_skinwrap selftest depth_scale_permil scales the whole table (1000 = the published depths; 0 = a pure copy of the bone). license_tier: ORIGINAL expect_exit: 0 No hw writes (Rule 26).