risk-adjustment

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No security issues detected. The skill provides functional code examples for data processing tasks consistent with its stated purpose.\n- [PROMPT_INJECTION]: The logic involves ingesting external healthcare data (CSV/DataFrames). This represents an indirect prompt injection surface; however, the skill performs standard data cleaning and the vulnerability is inherent to the intended data-processing use case.\n
  • Ingestion points: Crosswalk files and diagnosis datasets ingested via pandas in load_crosswalk and map_diagnoses_to_ccs.\n
  • Boundary markers: None present in the code templates.\n
  • Capability inventory: Data manipulation using pandas; no network or subprocess execution.\n
  • Sanitization: Code includes normalization for ICD-10 codes (period removal and case conversion).
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 06:00 PM
Security Audit — agent-trust-hub — risk-adjustment