ab-testing-ecommerce
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions and code snippets provide standard implementation patterns for A/B testing. References to external tools target well-known experimentation platforms and statistical calculators.
- [INDIRECT_PROMPT_INJECTION]: The skill describes processes for ingesting user identifiers and conversion data to perform experiment assignment and analytics tracking. This represents a standard surface for data ingestion in analytics applications, but the associated capabilities are limited to reporting and deterministic logic, posing no significant security risk.\n
- Ingestion points: User IDs (userId, customerId) and event data (order.total) mentioned in SKILL.md and the evaluation task descriptions.\n
- Boundary markers: Not present in the provided implementation templates.\n
- Capability inventory: Analytics event tracking and conditional logic for variant assignment.\n
- Sanitization: Standard handling of numeric and string identifiers as per common experimentation library usage.
Audit Metadata