These are header templates, not research data.

observation_metadata.csv
Add one row per spatial observation. id must exactly match the abundance matrix
and coordinate IDs. sample_id identifies an independent biological specimen;
all sections from the same specimen share that ID. condition identifies the
experimental group. Differential analysis requires real biological replication.

reference_signatures.csv
Replace CellTypeA / CellTypeB with actual reference cell-type labels. Add one row
per reference gene, using feature_id values matching the spatial RNA matrix.
Values must be count-scale cell-type mean expression or previously estimated
cell2location reference regression signatures. This table is derived from an
annotated single-cell reference, not invented from the spatial data. Spatial
input must contain raw, nonnegative integer RNA counts.

Upload the applicable completed template alongside the spatial data.
Alternatively, H5AD can store sample_id and condition in obs, and a gene-by-type
DataFrame in varm['reference_signatures']. For an array, also supply
uns['reference_cell_types'] in exactly the same column order.