generate-video-by-seedance
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and image references that are interpolated into API requests. This creates a vulnerability surface for indirect prompt injection. \n * Ingestion points:
scripts/generate_video.py(via command line arguments and YAML configuration). \n * Boundary markers: None present. \n * Capability inventory: Network communication (requests), local file reading (images), and file writing (videos). \n * Sanitization: No sanitization for LLM-specific instructions is performed on inputs. \n- [EXTERNAL_DOWNLOADS]: The skill downloads video files from ByteDance's official Volcengine Ark domains (volces.com). This is documented as the primary method for retrieving generated content. Evidence:scripts/generate_video.pyandscripts/get_video_task_status.pydownload from URLs returned by the API response.
Audit Metadata