bio-pathway-wikipathways
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- Indirect Prompt Injection (SAFE): The skill demonstrates a data ingestion surface via local file access. 1. Ingestion points: read.csv('de_results.csv') in SKILL.md. 2. Boundary markers: None present. 3. Capability inventory: write.csv (file writing) and downloadPathwayArchive (network download of data files). 4. Sanitization: None present. Analysis: The risk is categorized as SAFE because the tool is designed to process gene identifiers within a scientific R environment, making the execution of natural language instructions from the input data highly improbable.
- External Downloads (SAFE): The tool utilizes rWikiPathways to download pathway definitions (GMT files) from the WikiPathways repository. This is an expected and safe function for biological research.
Audit Metadata