seedance-2-0
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user prompts and external media assets which serve as ingestion points for indirect prompt injection. Ingestion points include text prompts and media URLs referenced in SKILL.md. Boundary markers are partially implemented through recommendations for explicit subject labels and quoted dialogue to separate data from instructions. Capability inventory includes making authenticated network requests to video generation APIs and using tools like ffprobe for asset validation. Sanitization measures include hashing internal identifiers for the safety_identifier field and normalizing media metadata.
- [EXTERNAL_DOWNLOADS]: The skill provides authoritative endpoints and schemas for interacting with well-known AI services such as BytePlus, fal.ai, and Replicate for remote model inference.
- [COMMAND_EXECUTION]: The documentation recommends using the ffprobe utility to technically validate media assets before transmission to remote APIs.
Audit Metadata