Version 0.1.0 is prepared but has not been published. Once released, the latest 0.1.x patch will receive fixes while that series is supported. The library accepts in-memory NumPy arrays; it does not deserialize files, load models, execute plugins, or access a network. Callers must still bound input size before processing untrusted arrays: voxelization allocates memory proportional to the input and sorts coordinates.

Do not post suspected vulnerabilities or credentials in public issues. Use GitHub's private vulnerability reporting form to contact the maintainer confidentially.

Include a minimal synthetic reproducer, Python/NumPy versions, expected behavior, and actual behavior. Avoid sharing private point clouds or other sensitive data in the report.