The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses ov.utils.load_metabolights to fetch public metabolomics datasets from the European Bioinformatics Institute (EBI) Metabolights public mirror at ftp.ebi.ac.uk. This is a well-known scientific data repository and the operation is documented as part of the skill's intended purpose for metabolomics analysis.
[DATA_EXFILTRATION]: The ov.metabol.run_mofa function generates a local HDF5 file to store multi-omics integration models. This is a standard persistence mechanism for large biological models and does not involve unauthorized data transfer.
[COMMAND_EXECUTION]: The ov.metabol.mixed_model interface utilizes Patsy formula syntax (e.g., 'treatment + time') to define statistical models. While this involves string-based model specification, it is the standard interface for the underlying statsmodels library.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data, such as metabolite names and database identifiers retrieved from the Metabolights database, and uses them in downstream analytical workflows and visualizations.
Ingestion points: Data is ingested via ov.utils.load_metabolights from external FTP sources in SKILL.md (Quick Workflow §16).
Boundary markers: Explicit boundary markers or instructions to ignore embedded instructions in metabolite names are absent.
Capability inventory: The skill includes capabilities for network data retrieval (ov.utils.load_metabolights) and local file writing (ov.metabol.run_mofa).
Sanitization: The skill relies on standard data frame processing and does not specify custom sanitization for external identifiers before interpolation into analytical routines.

omicverse-bulk-metabol-pathway-multifactor