medical-imaging-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [NO_CODE]: The provided file contains markdown documentation and metadata but no executable script or logic for analysis.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process DICOM metadata from external PACS servers. This represents a potential surface for indirect prompt injection if metadata (e.g., Study Descriptions) contains malicious instructions. Ingestion points: DICOM server responses (via C-FIND/QIDO-RS); Boundary markers: None documented; Capability inventory: Network access via DICOM/HTTP protocols (pynetdicom/requests); Sanitization: Not specified in documentation.
Audit Metadata