The Agent Skills Directory

[SAFE]: The skill implements legitimate scientific computing tasks using established libraries such as pandas, numpy, and scipy. The bundled script scripts/methylation_density.py and the code snippets in CODE_REFERENCE.md are focused entirely on statistical genomics analysis.
[SAFE]: No credentials or sensitive data are hardcoded; the skill correctly uses an .env.template file to manage external API keys for biological databases.
[SAFE]: Tool invocations are directed at reputable bioinformatics services (ENCODE, GTEx, Ensembl, etc.) and are consistent with the skill's primary purpose of biological data annotation.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection due to its processing of external datasets. 1. Ingestion points: Data is loaded from methylation CSVs, BED/narrowPeak files, and clinical metadata files in CODE_REFERENCE.md and scripts/methylation_density.py. 2. Boundary markers: No explicit prompt boundary markers or sanitization logic for input file content is present. 3. Capability inventory: The skill uses pandas for file I/O and calls external ToolUniverse APIs for gene and experiment annotation. 4. Sanitization: Input data is parsed into structured formats (DataFrames) using standard scientific computing libraries, which provides implicit type validation but no specific prompt injection filtering.

tooluniverse-epigenomics