AI Prompt Log — GodwulfNexus

A prompt log does two things a policy cannot. It shows a reviewer where another person's judgment was applied to AI output and where it was not, so the person checking the work knows where to look hardest. And it turns your own prompts into a way to test the tool itself. Run the same prompt again and the log shows you whether the AI gives you the same answer twice. Where it does not, you have found a place the output is unstable and needs a closer look. This is not a substitute for reviewing the work. It is a way to find out where reviewing matters most.

A high acceptance rate is not a sign the tool is good. It is often a sign the reviewer has stopped reviewing. If you are changing almost nothing, the question is not whether the AI is that good. It is whether you are still looking.

Where Your Data Goes

Nowhere. This tool runs entirely in your browser. When you add an entry, nothing is transmitted, because there is no server to receive it and no database to store it in. Your prompts and outputs never leave your device. When you close the tab, the log is gone, which is why you export before you leave. The exported file saves to your own machine and stays there.

You do not have to take my word for it. Open your browser's developer tools, go to the Network tab, and add an entry. You will see no request from this tool, because there is nothing to send. Any requests from your own browser extensions, a writing assistant like Grammarly, for example, are those extensions, not this page.

One note for full honesty: when the page first opens, it loads its spreadsheet-export library from a public code network, the same way it loads its fonts. That single request carries none of your data. After the page loads, nothing you type produces any request at all.

Want to Check for Yourself?

You can verify this claim, and check any other AI tool the same way, in about two minutes. I wrote a step-by-step guide to watching what a tool does and does not send from your browser.

Read the guide →

How to use this log

Add one entry per prompt. Record the AI tool you used and paste the exact prompt, so you can reuse or re-run it later.
For the output, you have two options. Write a one-line summary of what the AI produced, or paste the full output. The summary keeps your log scannable when you have many entries. The full output gives you a complete record. You can do either, or both.
Mark what you did with the output: accepted as-is, edited, or rejected. This is the most important field. Over time it shows your override rate.
Use the notes field to record what you changed and why.
When you are done, export the log to a spreadsheet. Nothing is saved until you export, so export before you close the page.

A note on the output field: the summary is there so a log with twenty similar prompts does not become twenty walls of text nobody will reread. Paste the full output when the exact wording matters. Summarize when you just need the record. The log works either way.