tooluniverse-target-research
Pass
Audited by Gen Agent Trust Hub on Mar 8, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill retrieves biological and chemical research data from numerous well-known and trusted scientific databases, including UniProt, Open Targets, the Protein Data Bank (PDB), and PubMed. These are established services in the research community.
- [PROMPT_INJECTION]: The skill features an indirect prompt injection surface due to its ingestion of large amounts of external literature and database descriptions. Ingestion points: Research abstracts and annotations retrieved via tools such as PubMed_search_articles and OpenTargets_get_publications_by_target_ensemblId. Boundary markers: Information is logically separated into structured sections using markdown headers as defined in the REPORT_FORMAT.md template. Capability inventory: The skill possesses network-access capabilities through predefined tools and file-system access for generating markdown reports. Sanitization: External text is incorporated into the reports without explicit sanitization; however, the risk of executable content in targeted scientific data is naturally low.
- [SAFE]: No evidence of direct prompt injection, obfuscation, privilege escalation, or unauthorized command execution was found during the analysis.
Audit Metadata