clinvar-variation-skill

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • External Data Retrieval: The skill interacts with government-hosted medical databases at NIH and NCBI. These are recognized sources for scientific information.
  • File System Interaction: The skill includes parameters for writing output to a local file path. This is a functional design choice for handling large datasets that would otherwise be truncated in chat, though it does grant the skill write access to the filesystem.
  • Indirect Prompt Injection Surface: By processing information from external databases, the skill is subject to the standard surface for indirect prompt injection common to all data-ingestion tools.
  • Ingestion points: External content is retrieved via HTTP requests in the clinvar_variation.py script.
  • Boundary markers: The skill does not specify delimiters for external data within the instructions.
  • Capability inventory: The skill can perform network operations and write files.
  • Sanitization: Structural validation is performed on the data, but natural language content is not filtered for agent-specific instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 12:20 PM
Security Audit — agent-trust-hub — clinvar-variation-skill