seedance-2-0
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill acts as a technical guide and set of templates for interacting with third-party video generation APIs (fal.ai, Replicate, Runway, etc.) within the OpenMontage framework.
- [CREDENTIALS_UNSAFE]: The skill correctly instructs users to manage authentication through standard environment variables (e.g., FAL_KEY, HEYGEN_API_KEY) rather than hardcoding sensitive secrets.
- [EXTERNAL_DOWNLOADS]: The skill references official documentation and platform endpoints for established AI services and technology providers, which are treated as legitimate resources for the skill's stated functionality.
- [INDIRECT_PROMPT_INJECTION]: The skill manages an untrusted data ingestion surface for the purpose of video generation. Evidence chain: 1. Ingestion points: user-provided prompts and multimodal reference assets (images, video, audio) in SKILL.md. 2. Boundary markers: utilizes structured prompt templates and identity-lock blocks to guide model output. 3. Capability inventory: includes Write access for saving generated .mp4 files and Bash for framework operations. 4. Sanitization: relies on external provider API safety filters for input validation. This surface is intrinsic to the skill's primary function and does not present an elevated risk.
Audit Metadata