## write_skeleton()


Write a geometry-complete, science-empty product and return its path.


Usage

``` python
write_skeleton(
    out_dir,
    product_id,
    *,
    dims,
    header,
    geometry=None,
    kernels=(),
    bodies=(),
    resolved_bodies=(),
    wavelength=None,
    subsampling=None,
    rings_in_fov=False,
    edge_on=False,
    overwrite=True
)
```


Every argument except the science arrays mirrors [cubegenpy.writer.build_hdulist()](build_hdulist.md#cubegenpy.build_hdulist), so a skeleton is a normal product in every respect but its data content.


## Parameters


`dims: Dims`  
Must be final. FITS allocates array space on write, so the science arrays stage 2 fills have to be the right size from the start.

`header: Mapping[str, object]`  
Primary-header values. `PIPESTAT` is set automatically and any caller value for it is ignored.

`wavelength: np.ndarray | None = None`  
Optional; NaN-filled if omitted, for the case where the wavelength solution is not yet known.

`subsampling: SubsamplingPolicy | None = None`  
The rule that chose `dims.NT`. Used to validate the value and to record its provenance in `NT_RULE`; defaults to `~cubegenpy.subsampling.SubsamplingPolicy`. `NT` itself is taken from `dims` - the policy does not override an explicit choice, it only refuses one outside its bounds.