skills/delorenj/skills/vox-tts/Gen Agent Trust Hub

vox-tts

Warn

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: MEDIUMPERSISTENCEDATA_EXFILTRATIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [PERSISTENCE]: The scripts/register_mcp.sh script modifies local application configuration files to register the service as an MCP server. It specifically targets ~/.hermes/config.yaml and ~/.claude/settings.json, which ensures the remote tool remains active across agent sessions.
  • [DATA_EXFILTRATION]: The scripts/clone_voice.sh and scripts/add_voice.sh scripts facilitate the upload of local files to a remote server (https://vox.delo.sh). While this is the intended functionality for voice cloning, the mechanism allows for the exfiltration of local data if the agent is directed to upload sensitive files instead of audio recordings.
  • [DYNAMIC_EXECUTION]: Multiple scripts (scripts/register_mcp.sh, scripts/synth.sh, and scripts/synth_url.sh) utilize python3 -c and `python3
  • <<EOF` to execute dynamic code blocks for JSON construction and configuration file manipulation at runtime.
  • [COMMAND_EXECUTION]: The skill relies on shell scripts that invoke several system commands including curl, python3, sed, tr, and CLI tools for other agents (like hermes and openclaw) to perform registration and synthesis tasks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 4, 2026, 06:55 AM
Security Audit — agent-trust-hub — vox-tts