modeling-conversion-metrics

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies event names and properties as untrusted ingestion data.
  • Ingestion points: Event names and property keys in the database events table.
  • Boundary markers: Instructions specify treating these as quoted data rather than instructions.
  • Capability inventory: Use of tool-based view creation for persistent storage.
  • Sanitization: Requires use of schema verification tools before modeling.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 10:15 PM
Security Audit — agent-trust-hub — modeling-conversion-metrics