The Agent Skills Directory

[SAFE]: The skill provides documentation and code examples for the scvi-tools library, which is a standard tool in the life sciences domain for probabilistic modeling of single-cell data. All documented external resources and dependencies are well-known scientific packages.
[INDIRECT_PROMPT_INJECTION]: The skill includes an attack surface for indirect prompt injection by processing untrusted data.
Ingestion points: Data is loaded via sc.read_h5ad, sc.read_csv in workflows.md, and sc.read_visium in models-spatial.md.
Boundary markers: None explicitly used in prompts, though the operating procedure requires manual validation of results.
Capability inventory: The skill performs file system operations such as model.save and adata.write in workflows.md.
Sanitization: Relies on the standard HDF5/CSV parsing of the underlying libraries.
[EXTERNAL_DOWNLOADS]: The skill references the installation of standard packages from official registries and the download of genomic sequences via scvi.data.add_dna_sequence, which is standard functionality for genomic analysis tools.

scvi-tools