One change, same AI.
Your prompt
Summarise {{DOC}} in five bullets, one per risk.
A stronger version
Saved prompt: risk-summary Template: Summarise {{DOC}} in five bullets, one per risk. Slot: {{DOC}} — the document being reviewed. Known failure: invents a fifth risk when fewer than five exist. Fix: state the count found before listing, and stop there if it's under five.
Here is the test: open one saved prompt. Could a stranger use it correctly, and would they know, before running it, the one way it's failed before?
Try it on your own work.
Prompting Wizard is an open source prompting coach for Claude Code, Codex, ChatGPT and Claude, built around your own work. Thirty guided lessons, at your own pace.