talking-head-production

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references installation instructions for the 'belt' CLI and documentation hosted on the official GitHub repository for inference.sh, which is a well-known service in the AI ecosystem.
  • [COMMAND_EXECUTION]: The skill documents the use of the 'belt' CLI to execute AI model workflows (such as pruna/p-video-avatar and bytedance/omnihuman-1-5) on a remote platform. The use of the Bash tool is scoped to this specific CLI tool as indicated in the allowed-tools metadata.
  • [PROMPT_INJECTION]: The skill provides examples of processing user-provided text via the 'voice_script' and 'voice_prompt' parameters. While these represent ingestion points for external data, they are used within the intended functional parameters of the generative AI models and do not attempt to bypass agent safety controls.
  • Ingestion points: User-provided strings for 'voice_script', 'voice_prompt', and 'video_prompt' in the belt CLI commands.
  • Boundary markers: Commands use JSON structured input for the --input flag, providing a clear boundary for data interpolation.
  • Capability inventory: The skill uses the Bash tool to execute the 'belt' CLI.
  • Sanitization: None explicitly mentioned; the skill relies on the underlying models and the inference.sh platform for content moderation.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 06:32 PM
Security Audit — agent-trust-hub — talking-head-production