seedance-product-video

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a shell script (ofox-video.sh) via bash to interact with an external video generation API. This script is located at a relative path in a sibling directory (../ofox-video-core/), representing a modular dependency on a core utility provided by the same author.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text descriptions and local image files to generate video content, creating a surface for potential injection.
  • Ingestion points: The --prompt and --frame-first-image parameters defined in SKILL.md accept external data.
  • Boundary markers: The skill instructions guide the agent to wrap user input into a specific, neutral, catalog-style prompt format.
  • Capability inventory: Shell command execution (bash) and network operations via standard binaries (curl, jq) are used to transmit data.
  • Sanitization: Input data is transmitted to the vendor's API where content moderation and safety filtering are performed server-side.
  • [SAFE]: Network operations are directed to the vendor's official domain (app.ofox.ai) for legitimate API functionality, and the skill correctly manages the OFOX_API_KEY environment variable for authentication.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 07:02 AM
Security Audit — agent-trust-hub — seedance-product-video