blog-audio

Fail

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The scripts/requirements.lock file specifies versions for several Python packages that are significantly ahead of the currently available versions on the official PyPI registry, such as certifi==2026.4.22, google-genai==1.73.1, cryptography==47.0.0, and urllib3==2.6.3. The use of non-existent high version numbers with associated hashes is a characteristic indicator of a dependency confusion or supply chain attack targeting automated build systems.
  • [REMOTE_CODE_EXECUTION]: The wrapper script scripts/run.py is designed to execute other Python scripts within the skill's virtual environment. However, it takes the script name as an unsanitized command-line argument and joins it to a local directory path without sufficient validation. This enables a path traversal vulnerability where an attacker could potentially execute any Python file on the system that the agent has access to by providing a manipulated path.
  • [COMMAND_EXECUTION]: The skill frequently uses subprocess.run to execute external tools like ffmpeg and pip. While these calls are implemented using list-based arguments, the combination with the path traversal vulnerability in the runner and the suspicious nature of the dependencies elevates the risk of these operations being used maliciously.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jun 13, 2026, 08:44 AM
Security Audit — agent-trust-hub — blog-audio