skills/alivirgo/major-ai-skills/horos/Gen Agent Trust Hub

horos

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses standard macOS and DCMTK command-line utilities like echoscu, open, defaults, and find for application control and network diagnostics within a local environment.
  • [EXTERNAL_DOWNLOADS]: References well-known and industry-standard Python libraries pydicom and pynetdicom for medical imaging tasks. These are established packages from reputable repositories.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes scripts that ingest DICOM metadata from external files. While this represents a data ingestion surface, the risk is negligible as the operations are limited to printing metadata and local network transfers.
  • Ingestion points: Metadata extraction in send_dicom_to_horos.py and audit_dicom.py via pydicom.dcmread.
  • Boundary markers: None present.
  • Capability inventory: Local file system read access for medical images; network operations (C-STORE/C-ECHO) restricted to 127.0.0.1.
  • Sanitization: Relying on standard pydicom parsing; no explicit validation of metadata strings before logging.
  • [DYNAMIC_EXECUTION]: Provides Objective-C templates for Horos plugins (PluginFilter). These are legitimate developer resources for extending the viewer's functionality and do not involve runtime execution of untrusted code without manual user compilation.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 11:43 PM