The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installation steps include fetching the pydeseq2 and anndata packages from the public PyPI registry. It also references the load_example_data utility, which downloads synthetic CSV files from the official PyDESeq2 GitHub repository managed by Owkin.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied count and metadata files for statistical processing. Although the provided driver script implements checks for numeric integrity and identifier uniqueness, the processing of metadata strings introduces an entry point for potential indirect instructions.
Ingestion points: The read_indexed_csv function in scripts/run_deseq2_analysis.py parses external CSV files using pandas.read_csv.
Boundary markers: The script validates the structural alignment of metadata and count indices but does not utilize explicit prompt delimiters to separate external data from agent instructions.
Capability inventory: The skill is capable of writing various file formats to the local disk (CSV, H5AD, and PNG) and performing network operations to manage dependencies and fetch bibliographic metadata from arXiv.
Sanitization: Input validation is present in the validate_counts function, which ensures data is numeric, finite, and contains only exact nonnegative integers.

pydeseq2