Most prompt libraries are graveyards. Prompts collected speculatively, never re-used, slowly aging into irrelevance. The valuable ones are different: the prompts you've already pressure-tested with real work.
Prompts as compound interest
Every prompt that produced something useful is a small piece of leverage. It captures a way of asking that worked. the role you framed, the constraints you imposed, the format you specified, the examples you supplied. That framing took effort to discover. Once discovered, it's reusable for a thousand future tasks of the same shape.
The math is brutal. A team that saves and reuses 50 strong prompts across a year is functionally faster than a team that re-derives them every time. The output difference compounds.
How to know which prompts are worth saving
A few heuristics that work in practice.
It produced something you actually used. Not something you generated and admired. something that ended up in a doc, a spec, a deliverable.
It would work on a different input. The role, format, and constraints generalize. You can swap the specific topic and the prompt still earns its keep.
You'd be embarrassed to lose it. A real signal. If discovering you can't find the prompt next month would frustrate you, save it now.
From prompt to skill
The next step up is a parameterized version. a Skill, in Shelvia's vocabulary. When you find yourself using the same prompt with small variations across projects, promote it. Mark the slots where the inputs change. Save the rest.
That graduation is where personal AI use becomes professional AI use.
