In the last post, we covered the three architectures for reusable LLM tools. That post prompted some great follow-up questions about the practical differences between these approaches, especially how using command-line (CLI) scripts as tools fits into the picture.