food-beverage-content-production
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted external data such as client briefs, product labels, and recipe specifications to generate prompts for image and video generation tools.
- Ingestion points: The skill explicitly instructs the agent to collect information from a "client brief", "product label", and "recipe spec" (SKILL.md, Operating stance).
- Capability inventory: The ingested data is used to construct complex prompts that are then passed to external image/video generation models (SKILL.md, Generated image and video prompt translation).
- Boundary markers: The skill provides a structured "Prompt pattern" and "Prompt components" section intended to organize inputs and separate factual data from stylistic instructions.
- Sanitization: The skill includes a "QA checklist" and explicit "Truth constraints" (e.g., "no invented claims", "no extra ingredients") to prevent the inclusion of fabricated or malicious content derived from the input data.
Audit Metadata