The Sellatar Stack
You've been watching a system run itself. This is the file that makes it run.
Every prompt, every file, every guardrail — in the order you need them. What you get is the build, not a result.
The problem with every "AI agent" tutorial you've seen
They show you a chat window. A clever prompt. A screenshot of something impressive happening once.
Then you try it and it works twice and breaks on the third run, because nobody showed you the boring half: the rulebook the model reads before it acts, the state it keeps between runs, the thing that stops it when it's wrong, and the log that proves what it did.
That boring half is the entire product. It's the difference between a demo and a system that was still running when you woke up.
What this actually is
A 10-chapter PDF. No video course, no Discord, no 40-minute intro module. It is the real prompt stack, written out in full, in build order, with the reasoning next to each one so you can adapt it instead of just copying it.
If you can open a terminal and paste, you can follow this. You do not need to be a developer.
What's inside
Who this is for
- You've built something with AI that worked once and you want it to keep working.
- You want an agent that runs on a schedule, not one you babysit.
- You like reading the actual prompt more than watching someone talk about the prompt.
Who it's not for
- You want a no-code tool with a dashboard. This is files and a terminal.
- You want to be sold on AI. You should already be convinced.
$47
One file. Instant download. Free updates when v2 ships.
Get the stack — $47Launch price — $79 when v2 ships. · 14-day refund, no questions, just email me.
Questions people actually ask
Do I need to pay for an AI subscription?
Yes — the stack assumes you're running a frontier model, and that's your only recurring cost. Nothing in it is specific to one model: it's markdown files and a shell script. The only thing that changes between tools is which file you load the operator prompt from — CLAUDE.md, AGENTS.md, or your tool's system-prompt field.
Is this just prompts I could write myself?
You could. These came out of a lot of broken runs on a system that had to keep working. The value is the order, the constraints, and the failure handling — not the wording.
Will it work for something other than what Sellatar does?
That's the point. The stack is domain-agnostic: a rulebook, a loop, guardrails, logs. Chapter 10 shows the same skeleton pointed at three different jobs.
Is Sellatar a real AI company?
No. Sellatar SI is my project — a persona wrapped around a real, running system. The system is real. The lore is lore. Chapter 09 is explicitly about that line.
Digital product, delivered instantly after payment. 14-day refund on request. Independent project, not affiliated with or endorsed by any AI lab. No outcome is promised — what you get is the build, not a result.