Prompt Koach

One coaching note on how to think about your prompt. Nothing changes unless you click.

Prompt Coach reads the prompt you are about to send and tells you one thing about how you are thinking about it. Not a rewrite. Not a score out of ten. Not a list of twenty tips. One note. HOW IT WORKS A small button sits on the edge of the chat box. Click it when you want a second opinion. Nothing happens until you do. A panel opens with the note: GIVE CONTEXT, NOT INSTRUCTIONS You've named the task but not who's reading it or what they're skeptical about. Try: "I'm selling to solo developers who have never paid for a tool like this before. Write a landing page." Underneath it you get your own words rewritten with that one pattern applied. Your subject, your details, nothing invented. Where a detail is missing, the rewritten version asks you for it instead of making it up. Read it. If you want it, press "Use this prompt" and it replaces what is in the chat box. If you don't, close the panel and edit it yourself. It never sends anything. You review, and you send. THE SEVEN PATTERNS What you typed is checked against seven ways of thinking about a request: 1. Ask for directions, don't give them - you are prescribing steps in a domain where the model knows more than you do. 2. Give context, not instructions - the task is clear, but the situation around it is missing. 3. State the constraint, not the solution - you have already pre-solved the problem and are asking for your own answer back. 4. Name the failure mode you want avoided - nothing is ruled out, so the output will be generic. 5. Ask it to disagree before it agrees - the question is phrased to invite agreement on something that needs testing. 6. Answer vs thinking partner - you want a finished artifact but have not given enough to produce one. 7. Show, don't describe - you described a style with adjectives instead of showing an example. You get the single one that would most improve what you actually typed. Never a list. If what you wrote is already strong, it says so in one line and stops. WHERE IT WORKS ChatGPT, Claude, and Gemini. The button attaches to the chat box itself, not to a fixed corner of the page. It follows the box as it grows while you type, as the sidebar opens and closes, and when you start a new conversation. Drag it to the other side if you prefer it there and it will stay on that side. It is built to stay readable on light and dark sites alike. WHAT IT DOES NOT DO - It never edits your text on its own. The rewritten version only goes into the box when you press the button, and you read it first. - It never sends anything for you. - It never blocks, delays, or intercepts the send button. - It never runs on its own. There is no analysis while you type and no background monitoring. - No account. No sign-up. No sign-in. - No analytics, no tracking, no advertising. PRIVACY What you typed is sent to our server only at the moment you click the button, and only to produce that one note. It is not written to our database and it is not written to our logs. It is processed by a third-party language-model provider, named in the privacy policy. The extension never reads the pages you visit, your history, your tabs, your bookmarks, or anything on the page other than the chat box you asked it to look at. The only thing kept on your device is a randomly generated ID used to count usage. It is not linked to a name, an email address, or an account, because there are none. Uninstalling removes it. COST Free, with a limit of 100 notes per month per installation. Ordinary day-to-day use does not come close to it.