tooluniverse-network-pharmacology
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied entity names and retrieves data from multiple external bioinformatics APIs, creating a potential surface for indirect prompt injection.
- Ingestion points: User input parameters such as
entityandsecondary_entity, along with structured data retrieved from external tools (e.g., OpenTargets, ChEMBL, DrugBank) as described in Phase 1 and 2 ofSKILL.md. - Boundary markers: No explicit delimiters are used to wrap external content, although the instructions mandate the use of English terms for tool calls.
- Capability inventory: Capabilities are restricted to data retrieval, bioinformatics network analysis, and report generation. No functions for arbitrary code execution, unauthorized file system access, or suspicious network communication were identified.
- Sanitization: The skill does not explicitly define sanitization or validation procedures for the data retrieved from external tools.
Audit Metadata