ncarnate reads legacy scientific data — HDF4/HDF-EOS2 granules (AMSR-E, MODIS) and netCDF3/netCDF4/HDF5 — and writes recompressed, CF-annotated netCDF4. It does two jobs: losslessly recompress netCDF/HDF5 (change compression, shuffle, or chunking without changing a stored value), and convert HDF4/HDF-EOS2 to netCDF4 with the geographic coordinates reconstructed so the output opens directly in modern tools (xarray, QGIS, Panoply). For HDF-EOS2 it parses the StructMetadata layer and rebuilds CF-standard coordinates: GCTP grid projections become CF grid mappings with 1-D x/y and 2-D lat/lon via pyproj, and swath geolocation is attached as CF coordinates with dimension-mapped geolocation interpolated through ECEF space. The correctness contract is data fidelity — only storage changes, never the science values, and every output is verified against its source before it replaces anything.