Improve my LinkedIn prompt
Paste a prompt you already use. Ten structural checks, each with the specific fix. They are pattern matches, not a model — which means they are fast, private and honest about their ceiling: they can tell you a prompt has no output schema, not whether it is aimed at the right person.
Here's one example run
A typical first-draft outreach prompt, checked. Yours will fail different ones.
5 of 10 checks pass — a typical first-draft outreach prompt.
The live version
What it checks and what it cannot
What makes a LinkedIn prompt bad?
Almost always one of six structural faults rather than bad wording: no role, so the model answers as a general assistant; no stated objective, so it aims at the topic rather than the outcome; no constraints, so you edit every output by hand; no output format, so you get prose you have to reformat; no rule against invented facts, so it supplies a statistic that does not exist; and no awareness of LinkedIn’s send limits, so it produces a plan the platform forbids. All six are visible from the text of the prompt, which is why a checker can find them.
Does it send my prompt anywhere?
No. The checks are pattern matches that run in your browser. Nothing is uploaded, stored or logged — the page does not have an endpoint to send it to.
Can it tell me if my prompt is good?
No, and it says so under the results. It can tell you that a prompt has no output format; it cannot tell you whether the prompt is aimed at the right person or whether the task is worth doing. A prompt can pass all ten checks and still be pointed at an ICP that does not buy.