mlx-fine-tuning

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). The SKILL.md explicitly instructs fetching/using community and HuggingFace models (e.g., "--model mlx-community/Qwen2.5-3B-Instruct-4bit" and "--hf-path Qwen/Qwen2.5-3B-Instruct" in the Workflow/Model Preparation sections), which are public, user-contributed artifacts the workflow downloads and loads—constituting untrusted third-party content that can materially influence downstream behavior.