nx_bodyfit_canon_gate.nx

nx_bodyfit_canon_gate.nx -- DOES CANON MODE ACTUALLY TRANSFORM THE CANON, AND DOES IT REFUSE WHEN THE GENOME DOES NOT DESCRIBE IT? nx_bodyfit grew a second genome decoder (per-part ring scaling), and a decoder that silently COPIED its input would pass every downstream check there is: the body still generates, the renderer still renders, the judge still scores, and the search still reports a best-of-generation. Nothing in the pipeline can tell a working transform from a no-op, which is exactly the shape of defect that survives review and ships. nx_bodyfit_canon_gate (no args) ★★★THE IDENTITY CASE IS THE STRONGEST INVARIANT AVAILABLE: every gene at mid-envelope maps to a per-mille scale of exactly 1000, so the emitted canon must be BYTE-IDENTICAL to the source -- every R row, every negative offset, every P row, or the decoder is lossy somewhere. ★★BUT BYTE-IDENTITY ALONE IS PASSED BY A TRANSFORM THAT DOES NOTHING, so it is paired with a BITE cell: the extreme genome MUST differ from the source while the identity genome MUST match. Either tooth alone is vacuous; only the pair proves the transform both WORKS and IS NOT A COPY. ★IT ADDRESSES THE ARTIFACT BY THE PATH THE ORGAN ANNOUNCES (canon_path=), never by reconstructing the filename from bodyfit.conf's envelope midpoints -- a reconstructed name breaks silently the moment any bound is edited, and a gate that tests the wrong file reports GREEN about nothing. exit 0 GREEN | 1 RED | 3 SKIP license_tier: ORIGINAL. No hw writes (Rule 26).