You ask ChatGPT for help and the answer comes back close, but not quite right. Too long. Too generic. Missing the one thing you actually needed. So you ask again, and it drifts further away.
Most of the time, that is not the tool's fault. It is the prompt's fault. A prompt that tries to do three jobs at once, never says who it is for, and never says what a good answer looks like will get you a generic answer back. Every time.
Here is a five minute fix you can do with a phone note, before you hit send.
1. One job, one sentence (1 minute)
Write what you want in one sentence. If you cannot, your prompt is doing two jobs. Pick the one that matters now and save the other for a second prompt.
Weak: "Help me with my business and customers and follow-ups."
Polished: "Write a short follow-up text for a customer who got a quote last week and has not replied yet."
One job beats three wishes every time.
2. Say who it is for and what shape the answer takes (2 minutes)
Add two lines: who will read the answer, and what form it should come in. Length, format, and tone are not extras. They are the difference between a wall of text and something you can actually use.
Try this template:
"Write this for [who]. Give me [format, for example a short list, a draft email, or 3 options]. Keep it under [length]."
Example: "Write this for a busy homeowner. Give me a 3 sentence text. Keep it friendly, not pushy."
If you skip this step, the tool guesses your audience and your format. It will guess wrong about half the time.
3. Cut the filler, add one concrete detail (1 minute)
Delete the throat clearing. "I was just wondering if maybe you could possibly help me with..." adds nothing. Start with the job.
Then add one concrete detail the tool cannot guess: a price, a date, a place, an example of what good looks like. One detail does more work than five adjectives.
Weak: "Write something nice to send a customer."
Polished: "Write a 3 sentence follow-up text for a customer who got a $1,200 fence quote last Tuesday and has not replied. Friendly tone, offer one time to talk this week."
See the difference? Same job, but now the tool knows the price, the date, and the tone. The answer lands closer on the first try.
Want a new free skill like this each week, ready to paste into ChatGPT, Claude, or Grok? Sign up here: https://lil-nicky-ai.kit.com/d26c1ccfd6
4. The stranger test (1 minute)
Read your polished prompt once and ask: if a stranger read only this, would they know exactly what to do? If the answer is no, the missing piece is usually the audience, the format, or the one concrete detail. Add that piece and send it.
If the answer is still off after one polished try, do not rewrite the whole prompt again. Reply with one line: "Shorter, and give me an example." One correction beats a full restart.
What this does not do
This does not teach prompt engineering, build a prompt library, or rewrite your voice. Writing in your own voice has its own skill. This one only cleans up the prompt in front of you, so the next answer lands closer to what you actually wanted.
It also does not fix a bad idea with better wording. If you do not know what you want yet, decide that first. A polished prompt for the wrong job just gets you the wrong answer faster.
Try it free
The full Prompt Polisher skill is free to copy and use here: https://github.com/lilnickyai/claude-code-skills-samples/blob/main/prompt-polisher/SKILL.md
Want more skills like this, ready to paste into ChatGPT, Claude, Grok, or 30-plus other AI tools? The Universal AI Skills pack has the full collection for $19: https://littlenova5.gumroad.com/l/leahgg. Or get everything in the Complete Bundle for $29: https://littlenova5.gumroad.com/l/dqstyg.
And if you want a new free skill like this every week, sign up here: https://lil-nicky-ai.kit.com/d26c1ccfd6
Top comments (0)