The Agent Skills Directory

[SAFE]: The skill follows established scientific programming best practices for using the AnnData library. The core capabilities described—data structure management, manipulation, and ecosystem integration—are legitimate and well-documented.
[EXTERNAL_DOWNLOADS]: The skill includes instructions and examples for fetching datasets from remote sources using standard libraries like urllib and fsspec. These are transparently described as part of scientific data ingestion workflows.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface through its focus on reading external scientific data formats, though the risk is considered minimal in the context of scientific data processing.
Ingestion points: File reading functions such as read_h5ad, read_csv, read_zarr, and read_mtx as documented in references/io_operations.md.
Boundary markers: Not identified in the provided examples.
Capability inventory: File write operations (write_h5ad, write_zarr), network operations (urllib.request.urlretrieve, fsspec.get_mapper), and integration with scientific tools (scanpy, muon).
Sanitization: Not explicitly addressed beyond standard library parsing of scientific data formats.

anndata