If you give your AI agent, the software that does work for you on your computer, a long task, and then every few minutes you come back to check, you are not alone.
You write in Spanish or Hebrew and the answer comes back in English, because halfway through it read files in English. You told it in its instructions file to always answer in your language, and that held at first. Could the problem be something other than the agent? An instruction is a request to the model, the brain behind the agent, and the model can forget it.
Then you paste the client proposal template you use every week and forget to fill in a slot. It burns a whole round guessing, and you are back in front of the window.
And you have no idea how much room is left in the conversation before it starts to forget.
It all started with an income and expenses table I wanted to build for myself in Excel. After I built a few versions in Excel, I moved to Google Sheets, and from there it kept growing.
While booking sales meetings by phone, I started building systems and automations, processes that run on their own, for business owners. I built a system for business forecast analysis, and it served the company's marketing and sales very well. That's where I realized I had knowledge every business needs. And a system that works for a business cannot depend on someone sitting there watching it.
Axel Jedlitschka
Hear a sound and see a notification in the corner of the screen when a job that ran at least a minute ends, with how long it took and in which folder.
Get a notification when the work stopped on an error, like an overloaded server or a dropped connection, so you can type "continue" once the network is back.
If you write in Spanish or Hebrew, get the answer in that language, even after the agent read English files, without asking again.
Know that a request with a slot you did not fill in is not sent, and see exactly what is missing.
See in one line how much of the conversation is used, yellow and then red as it fills up, and open a new one in time.
A hook is a small program that Claude Code, the agent this kit is built for, runs by itself at fixed moments. An instruction in the instructions file is a request to the model, and the model can forget it. A hook is not the model, so it does not forget. It runs every time, checks, and decides.
An instruction asks the model to remember. A hook does not forget, because it is not the model.
Calm kit · a kit by Axel Jedlitschka
Your agent tells you when it finishes, replies in the language you wrote in, and never sets off on a half-filled request.
ZIP · Skill + 4 hooks
What's in the ZIP
One ZIP file · one command to install
No cost
Four hooks, a skill that explains each one, a one-command install, ready settings in three languages, and a 7-page guide
Price: No costNo credit card, and no part is held back for payment
The kit is yours right away, at no cost. You only get marketing messages if you tick the box, and you can unsubscribe at any time.
As soon as you leave your details, a thank-you page opens with a button to download the kit. You choose the guide's language in the form: Hebrew, English or Spanish.
By Axel Jedlitschka
There is no purchase and no payment here, so there is nothing to cancel. You can withdraw your consent to marketing messages at any time.
You leave your details, and a thank-you page opens right away with a button to download the kit.
Yes. No credit card, and no part is held back for payment.
Yes. You need a Mac or Linux. Unzip the file you downloaded, and from the folder it creates run one command in the terminal: bash install.sh --all-hooks. From there everything works on its own, and tuning can be done by asking your agent in plain words.
No. The installer backs up the file before every write, never deletes or changes other hooks, and checks the file after writing. Uninstalling returns it exactly to what it was.
The hook tells the agent explicitly to keep any part you asked for in another language. And even in the worst case it asks only once per message, and then always lets the work end.
The hooks run in Claude Code everywhere, but the terminal shows the most. For other agents, the kit's AGENTS.md has a paste-ready block with the same rules, as an instruction rather than a hook.
From the kit folder run bash install.sh --uninstall. Only the kit's entries leave your settings, and the files move to a backup folder.
Your details are used to send you the kit and to follow up on it. Marketing messages are sent only if you ticked the consent box, and you can unsubscribe at any time. You can also ask to see, correct or delete your details.
You can keep coming back to the window every few minutes, or you can go answer email and come back when you hear a sound.
A kit to install and a 7-page PDF guide: four small hooks, a skill that explains every hook to your agent, and a one-command install, at no cost.
Price: No costNo credit card
The kit is yours right away, at no cost. You only get marketing messages if you tick the box, and you can unsubscribe at any time.
No cost
Get the kit
