Context compactor

What it is

Context compactor turns a long AI conversation into a short brief that a fresh session can carry on from. Every message in a chat goes back to the model with the next one, so a long thread costs more each turn. A brief lets you start clean and keep your place.

Who it suits

People who keep one AI conversation going for hours or days, and people who pass a thread on to someone else.

If your chats are short and start fresh each time, you do not need it.

Where to use it

In any AI chat, on any conversation you can copy.

• Any AI chat: where you paste the skill’s prompt and the thread
• ChatGPT, Claude or another chat app: a thread that has grown long and slow
• Claude Code: a session near its context limit, where file names and commands have to survive word for word

How to set it up

What you need first:

• Where it runs: any AI chat
• A key: none
• An install: none

1. Sign in to Mindket and press Get it on this listing.
2. In your Mindket account, View shows the prompt to read and copy. This page shows it too.
3. Keep it where you can paste it again: a saved prompt in your chat app, if it keeps them, or a note.

How to use it

Paste the skill’s prompt into an AI chat and put the conversation where it says {{input}}, or the stretch of it where the decisions were made. Then open a new chat and send the brief as your first message.

<the skill’s prompt>
<paste the conversation here, in place of {{input}}>

1. You paste a long chat in place of {{input}}
2. The model keeps the goal, decisions, facts and open items
3. A brief of at most 200 words comes back
4. You start a new chat with the brief

Example

You paste a long debugging chat about a login test that fails on every second run, with the skill’s prompt above it.

You get a brief of at most 200 words, in five parts:

File names, commands and code names appear as they were written, so the test’s file and the command that runs it come through unchanged.

For newcomers and experts

If you are new. An AI chat rereads the whole conversation every time you send a message. That is why long chats slow down and cost more. This skill writes the short version of where things stand. Try it on a chat you have kept going for a while, and see whether the brief alone gets a new chat up to speed.

For experts. It is one prompt with a hard cap of 200 words and five fixed parts. Names, commands and code identifiers are kept by instruction, not by parsing. The compression is lossy and the model decides what to drop, so the reasoning behind a decision may not survive. Keep the original thread if you might need to audit it. When a chat is longer than the model can read, paste only the stretch that holds the decisions.

Good to know

• What it keeps: only what is in the text you paste
• What it drops: a reason or a number left out of the brief is gone from the new session, so read the brief before you rely on it
• Privacy: it never reads the page you have open