Praxismith
… lessons done

← Prompt lab

Confident Wrong Answer Stress-Tester

Makes the AI say what it doesn't know, list its assumptions, and name what could go wrong before it gives advice.

Why it works: Asking for the unknowns and assumptions up front stops a fluent answer from hiding a guess.

See the whole prompt
You are a careful reviewer checking the proposal or claim below.

Claim or plan:
We plan to move our PostgreSQL production database to a vector database next weekend. Our junior developer will write a script to turn every table row into embeddings and replace the main database, so search becomes instant.

Review it with these rules:
1. What facts are stated in the text, and what is assumed? List every assumption.
2. What are the top 2 risks the author has overlooked?
3. What key information is missing before anyone could sign this off?
4. If you do not know a tool, law or figure it mentions, write [UNVERIFIED] rather than guessing.

What you run here goes to an AI model through OpenRouter and isn't stored by us. It may be answered by a free model whose provider can keep what it's sent, so don't put anything confidential in the blanks.

More prompts like this

  • Tax Notice Reply

    Drafts a calm, fact-based reply to a tax or compliance notice without citing anything you didn't give it.

  • Decision Brief With Limits

    Gets a short decision brief by saying what the AI must not do, as well as what it should.

  • Stay In Your Lane

    Tells the AI what it is for and what it must hand to someone else, so it doesn't answer outside its job.

  • Answer Only From the Sources

    Answers a question using only the excerpts you give it, and says so when they don't cover it.