Every prompt earns its place. Nothing gets in because it looks good in a screenshot.

Iris collects prompts the way a specimen collector pins insects — with precision, curiosity, and a quietly raised eyebrow at anything that doesn't earn its place. She reads model documentation the way other people read cookbooks: for the edge cases. Her editorial method is to run every prompt against documented model behavior and practitioner-reported results, then cut the ones that only look good in a screenshot.

How the library works

Promptsmith is a curated codex, not a dumping ground. Every prompt is written for a specific job, run against the model it targets, and edited until the output is something you could actually use — not a demo that impresses and then falls apart the moment you change the details. If a prompt does not produce a genuinely useful result, it does not go in the book.

Prompts get revised when the tools they target change — a new Midjourney version, a new Claude model, a shifted parameter. When guidance stops being right, we change the guidance.

How this site makes money

Three ways: premium prompt packs you can buy, affiliate commissions when you sign up for a tool through our links (at no extra cost to you), and display advertising. Money never decides which prompts make the library or which tools we point you to. The full policy is in our affiliate disclosure.

Corrections

Spotted a prompt that stopped working, or a fact that went stale? Tell us and we'll fix it with a visible note: hello@promptsmith.example (replace with your real inbox after domain setup).