Every week I see posts like "what are your favorite prompts?" followed by 50 generic suggestions, and then the same thread appears again two weeks later.
The problem isn't that people don't share prompts. It's that most prompt libraries aren't built for reuse — they're just dumps of things that worked once.
Here's how I built one that my whole workflow actually depends on.
The structure that makes a library usable
A prompt is only reusable if you document three things alongside it:
- The exact prompt (with placeholder variables in brackets)
- The use case (what specific output you need it for)
- The quality bar (what "good output" looks like so you know when to stop iterating)
Without #3, you don't have a library — you have a list of experiments.
Categories that cover 90% of knowledge work
The prompts that anchor each category
Email / edit_shorter.md:
Decisions / devils_advocate.md:
Copywriting / landing_page.md:
Operations / sop_builder.md:
Maintenance
Once per month:
- Review every prompt you used that week
- If you edited the output heavily, the prompt needs work
- If you used it 5+ times without editing, it's a keeper
Remove anything you haven't used in 2 months. A smaller, higher-quality library beats a comprehensive one you don't trust.
The full set I've built
I've been doing this for a year and have 47 prompts that have survived the monthly review.
Get the AI Prompt Pack () — organized exactly this way, by category and use case, with notes on when each one is appropriate.
The most useful part might be what's NOT in it — the 150 prompts I tested and cut because they didn't hold up.
Top comments (0)