video-generation-gateways
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes an integration pattern where user-supplied prompts and parameters are ingested and routed to third-party video generation gateways, creating an indirect prompt injection surface.
- Ingestion points: User prompts and shot design parameters are submitted to gateway endpoints (e.g.,
queue.fal.run,api.replicate.com) as detailed inSKILL.md. - Boundary markers: The documentation does not specify the use of prompt delimiters or explicit instructions to ignore embedded commands in the user-provided data.
- Capability inventory: The integration involves network requests to external inference APIs and the retrieval/download of hosted media results from CDNs.
- Sanitization: The skill notes that 'stacked moderation layers' (such as OpenAI's Omni moderation) are present but warns that safety behavior is inconsistent across different gateways and models.
Audit Metadata