ai-voice-cloning

Pass

Audited by Gen Agent Trust Hub on Jun 12, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references external repositories and documentation for CLI installation and extended functionality.
  • Directs users to install the 'belt' CLI using npx skills add belt-sh/cli.
  • Links to external installation documentation hosted at https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md.
  • Recommends adding related skills from the inference-sh/skills repository.
  • [COMMAND_EXECUTION]: The skill utilizes the belt CLI tool to execute voice generation and media processing tasks.
  • Provides numerous examples of executing belt app run with various AI models (e.g., infsh/kokoro-tts, inworld/text-to-speech-2).
  • Uses shell redirection to manage and store output files (e.g., > speaker1.json).
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it interpolates user-controlled data into tool commands.
  • Ingestion points: User-provided strings are directly embedded into the JSON payload of the belt command via the --input argument (specifically the 'prompt' and 'text' fields).
  • Boundary markers: The instructions lack explicit boundary markers or security delimiters to isolate user-supplied content from the execution context.
  • Capability inventory: The skill is authorized for Bash(belt *) operations, enabling command execution and file system writes based on generated output.
  • Sanitization: There is no evidence of input validation or sanitization to prevent malicious instructions embedded in the input text from influencing agent behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 12, 2026, 10:27 PM
Security Audit — agent-trust-hub — ai-voice-cloning