elevenlabs-voice-changer
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs users to install the
belt-sh/clitool and references installation guidelines hosted ongithub.com/inference-sh. These are functional dependencies for the skill's core purpose. - [COMMAND_EXECUTION]: The instructions rely on executing shell commands via the
beltCLI (e.g.,belt login,belt app run) to perform voice transformation tasks. - [INDIRECT_PROMPT_INJECTION]: The skill processes external audio data via URLs provided in the JSON input. This represents a potential surface for indirect injection if the processed content influences subsequent agent logic.
- Ingestion points: The
audiofield in JSON input blocks withinSKILL.mdaccepts external URLs. - Boundary markers: None explicitly defined in the instructions to separate untrusted data from processing commands.
- Capability inventory: The skill uses the
beltCLI tool to interact with remote ElevenLabs models and can pipe output to other tools (e.g.,media-merger). - Sanitization: No specific sanitization or validation steps are mentioned for the input URLs before they are passed to the CLI.
Audit Metadata