medical-imaging
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides guidelines for medical imaging interpretation (X-ray, CT, MRI, Ultrasound) and includes a Python script for categorizing findings. No malicious code, credential harvesting, or unauthorized network operations were identified.
- [METADATA_POISONING]: There is a minor inconsistency between the metadata author field ("HALO Core") and the identified skill author ("aizech"). This appears to be branding rather than a deceptive attempt to impersonate another entity.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external medical findings and DICOM metadata, creating a surface for indirect prompt injection.
- Ingestion points: The
analyze_findings.pyscript accepts a JSON file as input containing findings text. - Boundary markers: No specific markers or delimiters are used in the prompt instructions to isolate external data from system instructions.
- Capability inventory: The script performs local file reading and JSON parsing using standard Python libraries. No subprocess calls or network requests are present.
- Sanitization: No input validation or sanitization of the findings text is implemented before processing.
Audit Metadata