ai-voice-cloning

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the belt CLI tool to execute remote applications for text-to-speech and audio synthesis, including models from Inworld and ElevenLabs. Examples include commands such as belt app run and belt login.
  • [EXTERNAL_DOWNLOADS]: The skill references the inference-sh repository on GitHub for installation instructions and uses npx to download the belt-sh/cli tool. These references point to the service provider's infrastructure.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data via the prompt and text parameters in SKILL.md. This data is passed to external inference models. The capability inventory includes remote application execution and local file writing. No boundary markers or specific sanitization steps are defined within the skill instructions to mitigate instructions potentially embedded in the processed text.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 02:20 PM
Security Audit — agent-trust-hub — ai-voice-cloning