This is the way. I actually mapped out the decision tree for this exact process and more here:

https://github.com/NehmeAILabs/llm-sanity-checks