The Agent Skills Directory

[Dynamic Execution] (MEDIUM): The documentation for the 'DeseqDataSet.to_picklable_anndata()' method explicitly recommends using the Python 'pickle' module to save analysis results. As per security best practices, 'pickle' is inherently insecure because it can be used to execute arbitrary code during the deserialization process. This recommendation could lead users or agents to implement unsafe data loading patterns.
[Indirect Prompt Injection] (LOW): The API is designed to ingest external count matrices and metadata files through pandas DataFrames. This creates a vulnerability surface (Category 8) where untrusted data provided by an external source could potentially contain malicious content or be structured to manipulate the agent's interpretation of results, although the primary impact is limited to the statistical context of the tool.

pydeseq2