Prompt checker: is your request complete?
Paste a request you would send an AI agent or assistant into this prompt checker and see, as you type, which of the four parts of a good request it carries: context, task, guidelines and constraints. Each part found shows the words that gave it away; each part missing comes with a tip. It runs in your browser, and no AI model reads your text.
On this page
How to use the prompt checker
- Paste or write your request in the box, or try one of the two examples.
- Read the four rows. A part that is there shows the sentence it was found in, with the words marked.
- Add what is missing, using the tips, and watch the rows fill in.
- Check the vague words, if any: "better", "nice", "improve" leave the agent to guess what you mean.
How the prompt checker works (no AI model is called)
The checker cuts your text into sentences and looks, in each, for the signs of each part, in English, German and French:
| Part | What it is | Signs the checker looks for |
|---|---|---|
| Context | The situation, who it is for, the material given | "A customer wrote…", "for our sales team", "here is", "attached" |
| Task | What to produce | A verb asked of the agent: "write", "summarise", "draft a reply" |
| Guidelines | How: tone, format, structure | "friendly", "formal", "bullet points", "a table", "start with" |
| Constraints | The limits | "under 100 words", "at most 5", "don't", "never", "only" |
These are rules, not understanding: the checker finds the signs of each part, it does not judge whether the request is well written or whether its facts are right. A request can carry all four parts and still be unclear; a short one can be perfect for a simple job.
What the result means
A request with the four parts is one an agent can act on without coming back to you. Each part missing is something it has to ask about, or guess, and you then correct in another message. That costs more than time: an AI agent re-reads its instructions at every step it takes, about 20,700 tokens each time on the agent of our training sessions. In our sessions, "improve it" took 113,000 tokens where the precise request took 49,000.
Not every request needs every part. "Translate this into French" may need no guidelines. But when a result comes back wrong, the missing part is usually the reason.
To see what your request weighs in tokens, paste it into our token counter. To practise writing complete, lean requests, play Prompt Golf.
FAQ
Is the prompt checker free?
Yes, with no account. It runs entirely in your browser.
Does an AI read my request?
No. The checker is a set of rules run on your device: no AI model is called and nothing you type leaves the page.
What are the four parts of a request?
Context (the situation and who it is for), task (what to produce), guidelines (how to produce it: tone, format, structure) and constraints (the limits: a length, figures to keep, what not to do).
Why does the checker miss a part I wrote?
It looks for the words people usually use for each part. If you said it another way, it may not see it: that is the limit of rules. Writing the part plainly ("in a friendly tone", "under 100 words") usually helps the agent too.
Does a complete request guarantee a good answer?
No. It gives the agent what it needs, which makes a good answer far more likely on the first try. The facts, the examples and the clarity of what you ask still matter.
How do teams learn to write requests like this?
By practising on real work with a real agent and seeing what each request costs. That is what Tokenacademy's Prompting with AI program does. Book a demo to see a session.
See it with your own teams.
A 30-minute call: we get to know your project, then show you the platform.
Book a demo