wan-2-7
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text prompts and external URLs, which represents a standard indirect data ingestion surface.
- Ingestion points: The model accepts
promptandaudio_urlfields as data inputs inSKILL.md. - Boundary markers: Data is passed as a structured JSON payload via the
--inputargument to prevent accidental execution. - Capability inventory: The skill calls the local CLI tool
runcomfy run wan-ai/wan-2-7/text-to-videoto handle the generation process. - Sanitization: The CLI passes the input parameters directly as JSON over HTTPS without performing shell expansion, mitigating local command injection risks.
Audit Metadata