Do I have to finish all six modules?

No. Core 1 through 3, plus the takeover test on this page, is the whole course. Finish those three and you're done. Specialists 4 to 6 and the MCP page are optional extensions for the work that matches what you actually build.

Do I need Claude Code experience already?

Some. You should have it installed and have run a session before. Core 1 starts from a vague brief and teaches the prompting order from there, so you don't need to already know the method — just the tool.

Does this work with other AI coding tools?

The ideas transfer — write decisions down, review before building, keep one intent per session. The six prompts, the models named in the course (Opus scopes, Sonnet builds, Fable reviews) and the commands shown are Claude Code specific.

Is the reference project a real company project?

Tool Scout uses a fictional product team. The build process, checks and review findings come from a worked project; the business context and conversation excerpts are adapted for teaching. It does not describe a client or Bonanza's sales activity.

What do I get for my email?

One email unlocks every gated material across the course: skills, a hook and its wiring, sample dev-docs, a sample plan, and a prompt pack. Each module shows what's free to read and what's behind that one unlock.

How long does the core take?

Each core module runs on one focused session with Claude Code, plus the reading. Most readers finish Core 1 through 3 across a few sittings, not one sitting — the point is to actually run the method on your own project between modules, not just read about it.

Context Engineering for Claude Code

01

The curriculum

Six modules plus one standalone page. Everyone completes the same three core modules on the same project. Then you pick the specialist that matches your work.

Completing the course means Core 1 through 3, plus the takeover test below. Specialists are optional. Finish the core and you're done — the specialists just take the same project further.

Core — required

Core 1

Stop Claude guessing

Start from a vague one-paragraph brief. Learn the six prompts that separate what's known from what's assumed, before anything gets built.

Core 2

Stop starting over

Write the project down so a fresh session — or a teammate — can pick it up cold. Close the chat and find out what survives.

Core 3

Move from idea to evidence

Research before building, plan before coding, and get a second model to review the result before you call it done.

Specialists — optional, pick what fits

Specialist 4

Automate recurring work

Turn a procedure you've now done twice by hand into a skill, wired to a hook.

Specialist 5

Build in parallel

Run two agents on two worktrees at once, then reconcile what they built.

Specialist 6

Run research you can trace

One agent per claim, every source kept, gaps marked instead of guessed.

Standalone

MCP or a single API call

When a connector earns its place in the context window, and when a single API call is cheaper.

02

What context engineering for Claude Code means

Context engineering for Claude Code is the practice of putting everything Claude needs into files it reads on every session: the brief, the project instructions, the decisions already made and the state of the work. Prompting shapes one answer. Context engineering shapes every answer, for every person on the team, without the conversation that produced it.

Plenty of courses hand over Claude Code instructions and stop there. This one shows the instructions actually being followed — one project, adapted examples, the mistakes included — because the instructions are not the hard part. Keeping the work legible once the conversation ends is.

Every piece the course teaches answers the same question a different way. Where does this stand? Read the file names. What got decided, and why? Read lessons.md. How do two people work without blocking each other? A worktree per feature. None of it is a feature to learn for its own sake — it exists because a team loses sight of its own work the moment the person holding the context is unavailable.

“The work has to live in the repo, not in the person.”

The Operating Brain — the same argument at company scale makes the same argument at company scale, past what one repo can hold.

03

Can your teammate take over tomorrow?

This is the course's closing test, and it runs on your own project, not ours. Open your project's files — not your memory of the last conversation about them — and tick a statement only if the files themselves answer it.

Free, no email needed to see the score. The score is yours either way.

01 — GoalA teammate can find the goal — what we're building and why — in the files alone.
02 — Approved decisionsA teammate can find which decisions were approved, and which were rejected, in the files alone.
03 — Current stateA teammate can tell exactly where the build stands right now, in the files alone.
04 — Next actionA teammate can tell what to do next, in the files alone, without asking you.

Open your own project's files, then tick what they actually answer.

04