Hey HackerNews, excited to make my first post here. I run a small startup automating data flywheels for building performant Small Language Models. We made this Claude Code prompt which can help you identify if there are any LLM calls in your code where a frontier LLM is overkill or not suited for highly domain specific tasks. Thought I would post it here in case anyone might benefit!