tooluniverse-disease-research
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes extensive information from external bio-medical databases, creating a surface for potential indirect prompt injection where malicious content in these sources (e.g., publication abstracts) could influence the agent. * Ingestion points: Tool calls to PubMed, OpenAlex, OpenTargets, and other bioinformatics APIs in SKILL.md and TOOLS_REFERENCE.md. * Boundary markers: Data is structured into reports but lacks explicit instructions for the agent to treat external content as untrusted or to sanitize embedded instructions. * Capability inventory: Includes local file system writes and shell command execution. * Sanitization: No sanitization of retrieved external text is performed before inclusion in reports.
- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute a Python script (clinical_patterns.py) located in a related skill directory for differential diagnosis. While this is a cross-skill integration feature for the author's own tools, it utilizes shell command execution with parameters that incorporate user-supplied symptoms.
Audit Metadata