openevidence-core-workflow-b
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function involves processing external medical data and clinical scenarios, which could serve as a vector for indirect prompt injection if the source data contains malicious instructions.
- Ingestion points: The
queryparameter in the evidence search function and the metadata (titles, findings, summaries) retrieved from external medical databases as shown inSKILL.md. - Boundary markers: The instructions do not utilize specific delimiters or explicit 'ignore embedded instructions' directives when handling retrieved study content.
- Capability inventory: The skill allows the agent access to
Read,Write,Edit,Bash(restricted tonpmoperations), andGreptools. - Sanitization: There is no visible evidence of sanitization or validation of the data retrieved from medical search results before processing by the agent.
Audit Metadata