The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or persistence mechanisms were detected. The skill documentation describes legitimate use cases for bioinformatics data processing.
[EXTERNAL_DOWNLOADS]: The skill provides examples for downloading data from remote URLs using standard libraries such as fsspec and urllib. These references are for data ingestion purposes (e.g., fetching genomics matrices) rather than remote code execution.
[COMMAND_EXECUTION]: Installation instructions are provided for well-known, legitimate scientific packages (anndata, scanpy) using uv pip. No suspicious or unauthorized command execution patterns were found.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools for reading external bioinformatics data. 1. Ingestion points: ad.read_h5ad, ad.io.read_csv, and ad.io.read_zarr in references/io_operations.md. 2. Boundary markers: Not specified in documentation; assumes data follows library-specific schema. 3. Capability inventory: Filesystem read/write and python execution via allowed-tools. 4. Sanitization: Relies on the standard parsing logic of the anndata library. The risk is considered low as the data is used for scientific computation rather than instruction execution.

alterlab-anndata