Keep a Prompt Library Like a Chef Keeps Mise En Place
The best cooks never start from nothing, and neither should you. What I learned after a year of keeping every prompt that worked, organized by job instead of by tool.
Every professional cook works from mise en place: everything chopped, measured, and within reach before the heat goes on. Amateur cooks start each step from scratch, which is why their kitchen is chaos and dinner is late. Most people use AI tools the amateur way, improvising every prompt from zero, and after a year of keeping a prompt library organized like a kitchen station, I am convinced this is one of the largest unclaimed productivity gains available.
The library started embarrassingly: a note file where I pasted prompts that worked, because I was tired of rebuilding the same meeting-notes prompt from memory every Monday. The first lesson came fast: organize by job, not by tool. Prompts for summarizing, prompts for drafting, prompts for critiquing, prompts for extraction. Tools change monthly and models change quarterly; the jobs your prompts serve have not changed in years, and a library organized around tools becomes archaeology every time you switch.
A prompt in the library is not just the text; it is the recipe card. Mine carry three things: the prompt itself with its placeholders marked, one line on when it works and when it does not, and the date last verified. The verification line matters more than it sounds, because models change under you, and a prompt that produced gold in March producing cardboard in October is a real thing I have watched happen repeatedly. The date turns your library from a scrapbook into a maintained system.
The deeper value emerged over months: good prompts encode process, and a library lets you improve the process instead of the prompt. My blog-editing prompt is on its ninth version, each one incorporating what the last missed, and the diff between versions one and nine is essentially my editing philosophy written in executable form. When you rebuild from scratch every time, your best thinking evaporates with the chat window. The library makes your methodology compound.
Practical setup, for those who want one: any notes app works, mine is a markdown folder with one file per job; the overhead is minutes a week. The habit that makes it work is capture at the moment of success, pasting the prompt the second the output is good, because reconstructing a great prompt from memory two weeks later never reproduces it. Mise en place is not organization for its own sake, it is what lets a cook work fast under pressure. A prompt library does the same thing: the heat is on, the request is urgent, and your best thinking is already chopped and within reach.