The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of bioinformatics software from well-known registries, including PyPI, CRAN, and Bioconductor. It also references the mojaveazure/seurat-disk GitHub repository to support data conversion between R and Python formats.
[PRIVILEGE_ESCALATION]: The r_interop.md reference guide includes instructions for installing system-level dependencies (such as r-base, libcurl, and HDF5) using sudo apt-get or sudo dnf. These commands are standard for bioinformatics environment setup.
[INDIRECT_PROMPT_INJECTION]: The skill processes external biological data files, which creates a surface for indirect prompt injection if those files contain malicious instructions in metadata fields.
Ingestion points: The load_anndata function in scripts/_common.py reads data from .h5ad, .h5, .csv, .loom, and .mtx files.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are included in the processing pipeline.
Capability inventory: The toolkit provides file system write access and directory creation alongside complex data manipulation via the scanpy library.
Sanitization: The scripts do not perform explicit sanitization of cell metadata or gene identifiers during ingestion.

scanpy