skills/mbailey/voice-mcp/impressions/Gen Agent Trust Hub

impressions

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the 'voicemode' CLI to install a background service. This process creates a launchd unit ('com.failmode.voicemode.mlx-audio.plist') at login to maintain the local 'mlx-audio' server for voice synthesis.
  • [COMMAND_EXECUTION]: The skill relies on 'ffmpeg' for local audio manipulation, including trimming, normalization, and noise reduction. These operations are performed on local files within the user's home directory.
  • [EXTERNAL_DOWNLOADS]: The skill fetches the 'mlx-audio' package via 'uv tool install' and downloads AI model weights from Hugging Face (e.g., 'mlx-community/Qwen3-TTS'). These are established platforms for software and model distribution.
  • [PROMPT_INJECTION]: The skill has a data ingestion surface where it processes reference audio and transcripts.
  • Ingestion points: Reference WAV files and transcripts in 'voice.md'.
  • Boundary markers: None specified for the transcripts.
  • Capability inventory: Audio synthesis using local 'mlx-audio'. No direct system command or network write capabilities are triggered by the processed data.
  • Sanitization: None specified; output is restricted to audio.
  • [SAFE]: The skill references a companion repository (github.com/mbailey/voice-lab) which is a legitimate resource belonging to the skill's author.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 03:27 AM
Security Audit — agent-trust-hub — impressions