design-sound

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill performs audio processing and interaction with external APIs (ElevenLabs, Fal.ai) as described in its purpose. No malicious behavior or security risks were identified during the analysis.- [COMMAND_EXECUTION]: Multiple scripts, including mix.py, fit_music.py, fetch_web.py, video_foley.py, fal_foley.py, and ab_test.py, utilize subprocess.run to call external utilities like ffmpeg, ffprobe, yt-dlp, and nvidia-smi. These calls are implemented using list-based arguments and do not invoke a shell, which mitigates command injection risks.- [EXTERNAL_DOWNLOADS]: The skill includes functionality to fetch audio content from reputable and well-known sources, such as archive.org, incompetech.com, opengameart.org, and youtube.com. These operations are consistent with the skill's role as a sound design tool.- [DYNAMIC_EXECUTION]: score.py employs importlib to load a local project script (bank_samples.py), and video_foley.py dynamically modifies sys.path to interface with a local model installation. These are standard Python practices for managing local project dependencies.- [OBFUSCATION]: The automated detection of eval() in sound_library.py was determined to be a false positive. The code calls the PyTorch .eval() method to set an AI model to evaluation mode, which is unrelated to the dangerous Python eval() built-in function.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 12:28 PM