I build prompts that get used more than once. Before anything goes up for sale, I run it against a handful of real, different scenarios — not just the one example that happens to make it look good. If a prompt can't hold up outside its best-case demo, I don't list it.