It's tempting to set up folders, tags and a naming scheme before you've saved a single prompt. I'd skip all of that for now.
The prompt worth saving first is usually obvious once you look for it. It's the one you've reshaped three or four times for the same kind of job, where the last attempt finally came back the way you wanted. You'll need it again, and if it only lives in a chat you'll end up rebuilding it from memory and getting it slightly wrong.
That's the moment I built PromptAlo for, and it's free to add to Chrome: https://chromewebstore.google.com/detail/promptalo/jlfoipkmagdoafhlkhfkakdankpjhgni
What you save waits in the library, and one click sends any of it to ChatGPT, Claude or Gemini.
I've come round to thinking the first save shouldn't feel like a decision at all. You don't need a system on day one. You need a single prompt that's proven itself, saved. If the library grows from there, great. If it stays small because only a handful of prompts ever earned a spot, that's fine too.
Which prompt have you rewritten most often this month, and have you actually saved it anywhere?