advanced-customer-segmentation

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the ingestion of external customer data.\n
  • Ingestion points: The skill instructions refer to loading data from customer_orders.csv for analysis.\n
  • Boundary markers: There are no explicit delimiters or instructions provided to differentiate between data content and system instructions.\n
  • Capability inventory: The skill uses data for complex analytical tasks including Python-based K-Means clustering and life-cycle value calculations.\n
  • Sanitization: The provided files do not implement validation or sanitization of the input CSV content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 11:12 AM
Security Audit — agent-trust-hub — advanced-customer-segmentation