video-hook-first-3-seconds

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied scene descriptions and reference images to generate prompts for a video generation model.
  • Ingestion points: Scene descriptions and reference_images in SKILL.md.
  • Boundary markers: The skill uses structured fields in the generate_video call, though it lacks explicit "ignore instructions" safety delimiters.
  • Capability inventory: The skill uses a generate_video tool (Flux-3 model). It does not have access to sensitive local files, network exfiltration, or shell execution.
  • Sanitization: No specific sanitization or filtering of user-supplied text is described before interpolation into the prompt.
  • [EXTERNAL_DOWNLOADS]: The skill includes instructions to register an external Model Context Protocol (MCP) tool via mcp.apimage.org. This domain is recognized as belonging to the skill's author (apimageorg) and is used for its intended video generation functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 12:36 AM