dicom-metadata-extractor
Pass
Audited by Gen Agent Trust Hub on Apr 30, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses metadata fields from DICOM headers, such as Patient Name, Study Description, and Series Description. These fields are externally provided and could be manipulated to include malicious instructions that influence the agent's behavior.
- Ingestion points: DICOM files loaded from local or network directories for metadata extraction or anonymization processes.
- Boundary markers: No specific delimiters or instructions to ignore embedded content within DICOM tags are defined in the skill's instructions.
- Capability inventory: The skill has capabilities to read medical images and write structured reports (CSV/JSON) or modified DICOM files to the file system.
- Sanitization: While the skill's primary purpose is PHI redaction, it does not explicitly sanitize metadata values for potentially malicious natural language instructions before they are processed by the agent.
Audit Metadata