Skip to content

Check a prompt before you send it

Paste a prompt. You get a score out of 100, the weak spots, and how to fix each one.

Works for system prompts, agent tasks and workflows too, not just questions.

The weak spots and fixes are tuned to this goal.

0 of 5,000 characters

No prompt handy? Try one:

Draft prompt

Live preview
Write an email for a general audience.

Use a professional tone.

Format the output as plain paragraphs.

Target length: about 50 words.

Step 1 of 3: The basics

Paste a CLAUDE.md, AGENTS.md or .cursorrules file. You get a score, what it costs your agent every session, the rules it is missing and the installer text it carries.

0 of 40,000 characters

Free, counts as one of your 3 checks a day.

Example, for illustration

46out of 100Needs work before you send it

About 2,300 tokens, read at the start of every session.

About 1.2% of a 200,000 token context window, every session.

  • MajorMost of this file is installer text

    Three blocks pasted in by tools are read at the start of every session and say nothing about this project.

    In CLAUDE.mdAbout 1,800 tokens

  • CriticalNo rule about env files

    Nothing tells the agent which files hold secrets, so nothing stops it reading or committing them.

    In CLAUDE.md

Nothing found here.

  • MajorTwo rules about builds disagree

    One block says to run the build before every commit; your own rule says never to.

    In CLAUDE.md

Want your AI to stop re-solving the same problem? On Pro, you tell it to save what it worked out: the steps, the commands, how to check it worked. Next time, ask it to look that up in Claude Code.

See plans

How it works

  1. 1Paste a prompt

    Anything you are about to send to Claude, ChatGPT, Gemini or another AI.

  2. 2Read the weak spots

    A score out of 100 and what is missing, worst first, each with a fix.

  3. 3Take the rewrite

    Copy the improved prompt. A free account keeps it for you.

Questions

Is it really free?

Yes. You can run 3 checks a day with no account and no card. You see the score and every weak spot. The rewrite is free too; it needs a free account.

What does the score look at?

Seven things, the same way every time: how clear the goal is, who the answer is for, the format you want back, the rules and limits, the background the AI needs, the tone, and whether you show an example of a good answer. Each gets part of the 100.

What does a fixed prompt look like?

Here is a short prompt for a coding task, and the same request after the fixes.

Before

Write tests for this code.

After

Write Vitest tests for calcTotal() in src/utils/price.ts. Put them next to the file as calcTotal.test.ts. Cover an empty cart, one item, and a discount code that has expired. Match the style of src/utils/tax.test.ts. Each test should fail if the total is off by one cent.

Each added line answers one weak spot: which code, which framework and where the tests go, what to cover, what to copy, and what counts as done.

Which AI tools does it work with?

Claude, ChatGPT, Gemini and others. The fixes are about how the prompt is put together, so they carry across tools.

Do you keep my prompt?

Checks without an account are kept for 24 hours so your rewrite is waiting if you sign up. Signed-in checks are saved to your history. Do not paste passwords, keys or private customer data.

What do I get with an account?

The rewrite, a history of your checks, and deeper checks paid with credits.