review-generation-engine
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill designs a system that ingests untrusted user content through product review submissions.
- Ingestion points: The
POST /review/:tokenendpoint inreview-endpoints.tsis the primary entry point for user-submitted review titles and bodies. - Boundary markers: The design specifications in
tokenized-review-form-and-submission-end/criteria.jsonmandate character limits (100 for titles, 2000 for bodies) and minimum length validations. - Capability inventory: The
moderation.tsscript processes these reviews through a moderation pipeline that includes profanity checks, fraud detection (IP and frequency tracking), and regex-based content analysis before data is used to update aggregate product ratings. - Sanitization: Input is validated and truncated before persistence, and the moderation system filters out content that does not meet safety or quality criteria.
Audit Metadata