Skills
Navigate
skills/skills-101/superpowers/google-veo/Gen Agent Trust Hub

google-veo

Warn

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill directs the agent to download and install external CLI tools and configuration files from third-party repositories.\n
  • Evidence: Instructions to run npx skills add belt-sh/cli and references to https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.\n- [COMMAND_EXECUTION]: The skill's operation is centered on executing shell commands through the belt CLI tool to perform model inference.\n
  • Evidence: Multiple examples using belt login and belt app run with shell-executed arguments.\n- [PROMPT_INJECTION]: The skill ingests untrusted user-generated prompts and interpolates them directly into shell commands, presenting an indirect prompt injection surface.\n
  • Ingestion points: User-provided strings in the prompt field of the shell commands in SKILL.md.\n
  • Boundary markers: None present to delimit user input from command logic.\n
  • Capability inventory: The skill uses the Bash tool to execute commands with user-controlled input.\n
  • Sanitization: None. Input is passed directly to the belt CLI without validation or escaping.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 17, 2026, 04:09 PM
Security Audit — agent-trust-hub — google-veo