FOR CLAUDE CODE · CODEX · OPENCODE

The 12-course agent path. Free.The skill that wrote it. $4.99.

First prompt to multi-agent orchestration — in Chinese today, English on the way.

Get access · $4.99Read the path →
one sentence to your agentTeach me to put an AI agent to work properly.
✓reading your learner profile✓planning 5 lessons✓writing the exercises✓citing sources
learn-put-an-agent-to-workRenderedSource

Lesson 01: The Agent Is a Worker, Not a Talker

You already chat with AI. So why does handing it a real task still feel risky?

Most people start with AI the same way: they type a question and get an answer. You ask, "What is a good budget laptop?" and the AI answers in a paragraph. That is a chatbot. The AI stays passive. It responds to what you said, but it does not go off and do something for you. An agent is different. An agent takes a goal, chooses steps, uses tools, and returns a result. It can open a file, search the web, or run a small program for you.

the full lesson continues on your machine

saved to your machine · plain files

The next thing you want to learn probably isn't in those 12.

Those 12 courses are what one skill produced. The skill is the part you can own: install it in your own coding agent and it writes the next course from one sentence — the topic you actually need, pitched at the background you state.

  • Your topic, in your language
  • Plain Markdown, on your own machine
  • A local reader: practice graded on the spot, review at spaced intervals
  • Your own key, or hosted credits after you buy
$4.99
one payment · updates for at least 12 months
Get the skill

Reference price $4.99. The platform page has the final price.

what lands on your machine
written by your agent, from this skill and one sentence
01-worker-not-talker.md
# Lesson 01: The Agent Is a Worker, Not a Talker

> Lesson objectives:
> - Tell the difference between a chatbot question and an agent task.
> - Name the two things an agent needs before it can act: a clear goal and a way to check its work.
> - Rewrite one chat-style request into a task-style request.
>
> Prerequisites: None | Previous [<< README](./README.md) | Next [02 Write the ask >>](./02-write-the-ask.md)

## You already chat with AI. So why does handing it a real task still feel risky?

Most people start with AI the same way: they type a question and get an answer. You ask, "What is a good budget laptop?" and the AI answers in a paragraph. That is a chatbot. The AI stays passive. It responds to what you said, but it does not go off and *do* something for you. An agent is different. An agent takes a goal, chooses steps, uses tools, and returns a result. It can open a file, search the web, or run a small program for you[^S4].

Questions.

The path is the proof, not the product. Those 12 courses were written by this skill and stay free to read. What you buy is the thing that wrote them: install it in your own agent and it produces the same kind of course for the topic that is not on that list, pitched at the background you state, in the language you asked in.

A chat leaves you a transcript. This leaves you a course: lessons as files, sources you can open, practice graded locally, and review scheduled at spaced intervals. All on your machine.

Entirely on your machine: plain Markdown plus a local site. Put it in git if you like. Publishing a course to a shareable URL is optional, available after you buy Base. The lesson files stay on your machine either way.

Yes. A course is plain files, so it goes on any static host or your own server. After you buy Base you can also publish it as a URL from the local site: edit the course and the same link updates, with no server to run. Generating, reading, exporting, and reviewing locally stays free forever.

It can, and free prompts are everywhere. What you buy is the half a prompt cannot carry: a local runtime that grades practice, live review that writes fresh questions and returns them at spaced intervals, guard checks before a course ships, and the export chain to HTML, PDF, and Anki. That half is software.

For well-covered topics, a curated repo is good. Read it. The difference is elsewhere: here any topic becomes a course, pitched at the background you state, with practice graded on the spot and review that returns when due.

Three gates: a learner brief before writing, mandatory source records while writing, and 20+ course-guard checks before delivery. Every source is there to open and check yourself; a list that is all marketing copy or AI-generated fails guard outright.

Codex, Claude Code, and opencode are all verified out of the box. Any local agent that can read a skill directory and search the web can run it; without search, you can supply the sources.

Not for the core product. Producing, reading, exporting, and reviewing a course all run locally, for free. Three extras run on a model: in-place answers, web research while writing, and course audio. Bring your own key from any provider, at no charge to us, or use official hosted credits after you buy Base, billed per use with no subscription.

Your own agent is right for anything outside this lesson: it still owns deep work, updates, and everything past this lesson. Ask in place stays narrower. It answers against this course, this lesson, and whatever passage you select, so you get a short answer without leaving the page.

Yes. The reader is designed for absolute beginners, and one of the samples has no code at all. Topics follow your request — writing, spreadsheets, any craft; code exercises appear only in programming topics.

The charge appears as “PADDLE.NET* DFDIGITAL” on your bank or credit-card statement, because Paddle is the merchant of record for the purchase.