> ## Documentation Index
> Fetch the complete documentation index at: https://simpai.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Your sanitized prompt: what changed and why

> S.I.M.P. rewrites your prompt to remove vague language and add missing context. The changes panel explains exactly what was edited and the reason why.

After S.I.M.P. analyzes your prompt, it produces a sanitized version — cleaned and restructured to reduce hallucination risk. Just like sanitizing your hands removes germs, sanitizing your prompt removes vague language, missing context, and assumptions that confuse AI. The sanitized prompt is what you should actually use when you go to ChatGPT, Claude, or Gemini.

## The sanitized prompt panel

The sanitized prompt appears in its own panel below the risk score and issues list. It shows the full rewritten text, ready to copy. S.I.M.P. doesn't abbreviate or summarize the rewrite — you see the complete prompt, exactly as it would be submitted to an AI.

## The changes log

Below the sanitized prompt, a changes log breaks down every edit S.I.M.P. made. Each entry has two parts:

* **What:** a plain description of what was changed in the prompt.
* **Why:** the specific reason that change reduces hallucination risk.

For example:

| What | Why |
| - | - |
| Added output format specification ("respond as a numbered list") | Without a specified format, the AI may structure the response in a way that doesn't fit the use case, leading to follow-up prompts or misuse of the output. |
| Replaced "the system" with "the user authentication service" | Vague references to unnamed components force the AI to guess what's being discussed, increasing the chance of invented or incorrect details. |

Reading through the changes log is one of the fastest ways to build intuition for what makes a strong prompt.

## Before and after

Here's an example of how S.I.M.P. transforms a weak prompt into a specific, well-scoped one:

<Tabs>
  <Tab title="Before">
    ```text theme={null}
    Explain how authentication works.
    ```

    **Risk score:** 8 (High)

    Issues: no specified system or context, no target audience, no output format, no scope constraints.
  </Tab>

  <Tab title="After">
    ```text theme={null}
    Explain how JWT-based authentication works in a Node.js REST API. Assume the reader is a junior developer who understands HTTP but has not worked with tokens before. Structure your response as: (1) a one-paragraph overview, (2) a step-by-step flow of a login request, and (3) a bullet list of common mistakes to avoid. Keep the explanation under 400 words.
    ```

    **Risk score:** 2 (Low)

    Changes: added system context (JWT + Node.js), defined the audience, specified output format and structure, added a length constraint.
  </Tab>
</Tabs>

## How to use the sanitized prompt

Once you have the sanitized prompt, you have four options:

* **Copy** — copies the sanitized prompt to your clipboard.
* **Open in ChatGPT** — copies the prompt and opens [chatgpt.com](https://chatgpt.com) in a new tab. Paste and submit.
* **Open in Claude** — copies the prompt and opens [claude.ai/new](https://claude.ai/new) in a new tab. Paste and submit.
* **Copy for Gemini** — copies the prompt to your clipboard so you can paste it into Gemini manually.

<Tip>
  Use the "Run Again" button to load the sanitized prompt back into S.I.M.P.'s input field. This lets you run another analysis on the rewritten prompt and see how much the risk score has dropped — or catch any remaining issues.
</Tip>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.