pacs-workflow

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external PACS APIs, creating a surface for indirect prompt injection.\n
  • Ingestion points: Data enters the context via requests.json() calls in functions like orthanc_find_studies, dcm4chee_find_studies, and query_worklist in SKILL.md.\n
  • Boundary markers: None. The skill does not include delimiters or instructions for the agent to ignore or sanitize embedded instructions within the retrieved metadata.\n
  • Capability inventory: The skill utilizes the requests library for network operations and the open function for writing study data to the local filesystem.\n
  • Sanitization: No validation or sanitization of the retrieved DICOM tags (e.g., PatientName, StudyDescription) is performed before the data is processed or presented.\n- [DATA_EXFILTRATION]: The skill performs network operations to dynamic, potentially non-whitelisted PACS server URLs using the requests library. It also accesses and retrieves sensitive Protected Health Information (PHI) in the form of DICOM archives and metadata, which are subsequently saved to the local filesystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 11:46 PM