p-video-avatar

Pass

Audited by Gen Agent Trust Hub on Aug 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the belt CLI (belt app run pruna/p-video-avatar) to process data and generate videos. This is the intended purpose of the skill using the allowed Bash tool.
  • [PROMPT_INJECTION]: The skill supports the parameter disable_safety_filter: true, which explicitly instructs the backend to bypass content safety guidelines. While this is a feature of the underlying model, it represents a risk for generating restricted content.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external URLs (image and audio parameters).
  • Ingestion points: SKILL.md documentation and example commands show ingestion of external image (https://portrait.jpg) and audio (https://speech.mp3) files.
  • Boundary markers: None present in the command templates.
  • Capability inventory: The skill executes shell commands via the belt CLI.
  • Sanitization: No sanitization is mentioned; the skill relies on the backend API to handle potentially malicious payloads embedded in media metadata or content.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 12, 2026, 10:39 PM
Security Audit — agent-trust-hub — p-video-avatar