The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install mlx, mlx-lm, and mlx-vlm via standard package managers and provides examples of loading model weights from HuggingFace, which is typical for machine learning workflows.
[COMMAND_EXECUTION]: Provides examples of using local command-line tools for LLM generation, chat, and model conversion. These operations are restricted to the local development environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to handle user-supplied prompts and training data. Ingestion points: Processes external text via inference prompts and reads training data from local files (SKILL.md). Boundary markers: Demonstrates using tokenizer.apply_chat_template to manage chat boundaries (SKILL.md). Capability inventory: Includes fetching remote models, reading local datasets, and executing local binaries. Sanitization: The documentation focuses on framework usage; developers should implement appropriate sanitization for untrusted inputs in production.

mlx-apple-silicon