tooluniverse-clinical-trial-design
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and synthesize large amounts of untrusted external data from clinical registries and medical literature, creating a vulnerability surface where malicious content in those sources could influence agent behavior.
- Ingestion points: The skill uses tools such as
PubMed_search_articles,search_clinical_trials, andFAERS_search_reportsto retrieve external data from public databases (NIH, FDA, and academic repositories) into the agent's context. - Boundary markers: The skill does not define clear delimiters or "ignore embedded instructions" markers when interpolating external tool outputs into the feasibility report templates.
- Capability inventory: The agent has the capability to write files (creating the
.mdreport) and execute complex reasoning chains based on the ingested data. - Sanitization: There is no evidence of sanitization or validation logic to filter out potentially malicious instructions embedded in PubMed abstracts or trial descriptions before they are processed by the agent.
Audit Metadata