voice-isolator

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download the ElevenLabs CLI tool, Python library, and Node.js SDK from official vendor sources on GitHub and standard package registries.
  • [REMOTE_CODE_EXECUTION]: Includes a one-liner installation script (curl | sh) hosted on the official ElevenLabs GitHub repository for installing the CLI tool.
  • [COMMAND_EXECUTION]: Provides commands for installing dependencies using standard package managers (npm, pip, brew, scoop) and executing the elevenlabs CLI tool to perform audio isolation tasks.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes functionality to fetch and process audio from external URLs, representing an ingestion point for untrusted data.
  • Ingestion points: The audio_url parameter in SKILL.md used with the requests.get method.
  • Boundary markers: None present, as the data processed is binary audio rather than text instructions.
  • Capability inventory: The skill performs network requests to the ElevenLabs API and performs local file system write operations to save the isolated audio.
  • Sanitization: Not applicable for binary audio data in this specific use case.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 10:00 AM
Security Audit — agent-trust-hub — voice-isolator