nx_filehash.nx

nx_filehash.nx -- SHA-256 OF A FILE, AND THE ONE-CALL "ARE THESE TWO ARTIFACTS THE SAME BYTES" TEST. ★WHY THIS EXISTS, and it is a gap the ecosystem's own laws made expensive. This codebase carries a standing rule -- VERIFY THE INSTRUMENT AND THE ARTIFACT -- and a banked correction where a session claimed two meshes were "byte-identical", was wrong, and recorded the reason: "I inferred byte-identity from a matching tri count and matching bench scores and NEVER HASHED THE FILES." The reason nobody hashed the files is that there was no way to: nx_sha256 exists as a LIBRARY with a dozen gates and wasm builds, and `sha256_digest(bytes,n,out)` has been sitting in it the whole time, but NO ORGAN exposed it over a PATH. So every "byte-identical" claim in this program has been rhetorical. ★This is the adoption gap in its purest form: the primitive was built, gated and never wired to the one caller shape that every other lane needs. It is 90 lines. nx_filehash <path> -> {"sha256":"<64 hex>","bytes":N} nx_filehash cmp <a> <b> -> {"identical":0|1, both digests, both sizes} nx_filehash selftest -> KAT + non-vacuity teeth license_tier: ORIGINAL expect_exit: 0 No hw writes (Rule 26).