procedure-coding
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFE
Full Analysis
- External Data Processing: The skill is designed to analyze clinical notes, which are untrusted inputs. This creates a surface for potential indirect prompt injection. The skill manages this by limiting its output to standardized codes and relying on specific tool calls rather than open-ended text generation. Evidence Chain: (1) Ingestion points: Clinical encounter documentation mentioned in SKILL.md; (2) Boundary markers: None present; (3) Capability inventory: search_codes, validate_code, cpt_search_codes, and cpt_lookup_code; (4) Sanitization: None described.
- Use of Medical Connectors: The skill interacts with hcpcs_codes and cpt_codes MCP connectors. These are specialized tools for clinical data retrieval and do not provide the skill with access to the host file system or general internet connectivity, ensuring the skill operates within its intended medical scope.
Audit Metadata