Prompt Quality Checker

What it is for

A prompt can read well and still be underspecified. The gaps are hard to see in your own writing, because you know what you meant — the model does not, and it will not ask.

This grades a prompt against six dimensions and returns specific fixes rather than a verdict. The score matters less than the list of what is missing.

How it works

The checker looks for a stated role, sufficient context, an unambiguous task, a specified output format, explicit constraints, and language precise enough to act on. Each dimension gets a score and, where it falls short, a concrete suggestion.

It is a heuristic, not a model call. That makes it instant and private — nothing you paste leaves the browser — but it means it judges structure rather than meaning. A prompt asking for something impossible in a well-structured way will score highly.

When to use it — and when not to

Best used before a prompt goes into anything repeated: a workflow, an automation, a template you will hand to a team. A prompt run once can be fixed by looking at the output; a prompt run ten thousand times cannot.

Less useful for exploratory prompting, where vagueness is sometimes the point — if you are thinking out loud with a model, a low score is not a problem to solve.

Things worth knowing

Two dimensions do most of the work. Output format is the one people skip most often and the one that causes the most downstream pain, because a model that returns prose where you needed JSON breaks whatever consumes it. Constraints are the second — the difference between "summarise this" and "summarise this in under 100 words, without adding anything not present in the source" is most of the quality gap.