p-video-avatar
Warn
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: MEDIUMPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The parameters documentation includes a 'disable_safety_filter' option which is set to 'true' by default, representing an explicit safety bypass instruction.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted data in multiple parameters.
- Ingestion points: Input parameters 'voice_script', 'video_prompt', and 'voice_prompt' in SKILL.md.
- Boundary markers: Absent; user or external text is interpolated directly into the CLI command without delimiters.
- Capability inventory: Shell command execution via the 'belt' CLI tool to generate remote content.
- Sanitization: Absent; no escaping or validation is performed on the input strings before they are passed to the subprocess.
- [EXTERNAL_DOWNLOADS]: The skill directs users to download and install a CLI tool from a remote repository and references installation scripts hosted on GitHub (inference-sh/skills).
- [COMMAND_EXECUTION]: The skill's primary function is the execution of 'belt' shell commands to interact with the Pruna model API.
Audit Metadata