Prompt Grader& Fixer
Paste a prompt. Get an instant score across the five things that make prompts work — then a rewritten version with the missing pieces filled in. Free, no account, and it never leaves your browser.
Type or paste any prompt above and it’s graded instantly — no account, nothing leaves your browser.
The five dimensions
A good prompt is not about magic words — it is about giving the model enough to work with. The grader looks at the five dimensions that, again and again, separate a prompt that returns something useful from one that returns mush. Each is worth 20 points.
Clarity
Does the prompt open with a clear action — write, summarize, analyze — or does it wander? Vague openers like “help me with” or “do something about” leave the model guessing at the actual job.
Context
Has the model been given a role, an audience, and the background it needs? “You are an expert onboarding manager writing for new hires” produces sharper output than a cold instruction with no framing.
Specificity
Concrete beats generic. Names, numbers, the real subject matter, and a worked example push the model away from bland, hedge-everything answers toward something you can actually use.
Constraints
Length, tone, and do/don’t rules keep the response on target. “Under 150 words, friendly tone, no jargon” is the difference between a usable draft and an essay you have to trim.
Output format
Telling the model how to shape the answer — a numbered list, a table, three short sections — saves a round of reformatting and makes the result easier to scan and reuse.
Common questions
How does the prompt grader score my prompt?
It checks your prompt against five dimensions that consistently separate strong prompts from weak ones — clarity, context, specificity, constraints, and output format — and awards up to 20 points each for a 100-point total. The score is a guide, not a verdict; the real value is the specific tip under each dimension telling you what to add.
Does my prompt get sent anywhere?
No. The grader runs entirely in your browser using a transparent set of rules. Nothing you paste is uploaded, stored, or sent to any server or AI model. You can even use it offline once the page has loaded.
What does “Apply fixes” do?
It rewrites your prompt by adding the scaffolding it was missing — a role, an audience, length and tone constraints, and an output format — using [bracketed] placeholders you fill in with your specifics. You can copy the result or open it straight in ChatGPT or Claude.
Is a higher score always better?
Usually, but use judgement. A quick brainstorming nudge does not need every constraint nailed down. The grader rewards prompts that are explicit and self-contained, which matters most when you want a reliable, repeatable result.
Once your prompt scores well, browse the full library for ready-made, customizable prompts — or explore the rest of the Toolkit.