tooluniverse-gwas-drug-discovery
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface by processing untrusted external data.
- Ingestion points: The skill ingests data from two primary sources: user-provided disease traits in the
discover_drug_targetsandfind_repurposing_candidatesfunctions, and external scientific data fetched from APIs such as the GWAS Catalog, OpenTargets, ChEMBL, and the FDA. - Boundary markers: There are no clear delimiters or system instructions (e.g., "ignore embedded instructions") present in the skill's prompts or code to distinguish between control instructions and data retrieved from external bioinformatics sources.
- Capability inventory: The skill has the capability to interact with numerous external databases through the
execute_toolwrapper, including GWAS Catalog, OpenTargets, ChEMBL, DGIdb, and FDA adverse reaction databases. - Sanitization: No evidence of input validation, escaping, or sanitization is performed on user-provided disease names or the data returned from API tool calls before they are integrated into the agent's context for ranking and reporting.
Audit Metadata