parakeet

Warn

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [Unverifiable Dependencies & Remote Code Execution] (MEDIUM): The skill instructs the use of 'uvx parakeet-mlx', which fetches and runs a package from PyPI at runtime. Since PyPI is not a trusted repository under the [TRUST-SCOPE-RULE], this constitutes unverified remote code execution.\n- [External Downloads] (LOW): The skill downloads a ~2.5GB model file. Although Hugging Face is a trusted organization, the download of unverified large binary assets is a noted risk factor. Finding is downgraded per [TRUST-SCOPE-RULE].\n- [Indirect Prompt Injection] (LOW): The skill possesses an indirect injection surface (Category 8) by processing untrusted audio files. The severity is low because the skill produces text output (SRT/TXT) rather than using the content for downstream decision-making or command execution.\n- [Command Execution] (LOW): The skill utilizes local subprocess calls for utilities such as 'ffmpeg', 'awk', and 'grep' to perform audio conversion and text processing.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 16, 2026, 10:47 AM