Use Claude Code without losing control

Claude Code can inspect a repo, edit files, run commands, and prove work. This track teaches when to use it, how to brief it, where repo memory belongs, and how to control permissions, hooks, MCP, skills, subagents, and headless runs.

Start Claude CodeLesson 1 takes about 14 minutes

15: Lessons
3-6: Levels
4 free: Intro
11 paid: Advanced

Who this is for

People new to Claude Code

You know Claude chat, but repo work still feels like a separate world.

Repo owners

You want AI help without letting it touch auth, billing, data, or secrets by accident.

Teams standardizing AI work

You need shared instructions, review rules, and proof so sessions do not drift by person.

Power users

You are ready for hooks, MCP, skills, and subagents, but only when the setup earns its place.

Creative operators

You need code-agent help for sites or tools, but you still want visible proof before acceptance.

Automation builders

You want headless runs or SDK paths after the manual process has already worked.

4 Modules | 15 Lessons | Build in order

Module 1

4 lessons

Level 3 | 4 free lessons

Start safely

Use Claude Code only when the repo needs action, and undo without panic when it drifts.

This module teaches where Claude Code fits, which surface to start from, how to run a first session that inspects before editing and proves the work before acceptance, and how to rewind a session that goes sideways.

Proof:You can classify a repo task, choose the surface, write a safe first-session brief, and rewind a session back to a known-good point.

New Claude Code usersRepo ownersTool-switchers

Start Module 1

Module 2

4 lessons

Level 4 | 4 paid lessons

Set repo context

Put instructions, settings, and task facts in the right place, then sort the task itself.

This module covers CLAUDE.md, the .claude folder, local settings, shared rules, clean session packets, and the recipe-versus-chef call: whether a task needs fixed steps or an open-ended goal.

Proof:You finish with repo instructions, setup placement rules, a clean context packet, and a real task sorted into a recipe or a chef.

Team leadsPower usersCourse authors

Start Module 2

Module 3

3 lessons

Level 5 | 3 paid lessons

Control action

Use permissions, review, and hooks before trust grows too loose.

This module teaches allow, ask, deny rules, diff review, proof checks, and hook decisions. The goal is controlled speed with clear stop points.

Proof:You finish with a permission map, review checklist, and one hook decision.

Repo maintainersReviewersAutomation owners

Start Module 3

Module 4

4 lessons

Level 6 | 4 paid lessons

Extend only when needed

Add MCP, skills, subagents, and headless runs only after proof, then verify by risk, not habit.

This module keeps advanced setup tied to a real need. You add one connection, role, or automation path only when the task gap and proof justify it, then close the track by matching verification shape to what each task actually risks.

Proof:You finish with an MCP gap test, extension choice, automation readiness gate, and a verification contract for one real task.

Advanced operatorsTool buildersTeam system owners

Start Module 4

Start with one safe session

Claude Code should change files only after the task, boundary, and proof are clear. Lesson 1 helps you decide whether Claude Code owns the work before any command runs.

Claude Fundamentals Codex Fundamentals

Start Lesson 1

Use Claude Code without losing control

Start Claude CodeLesson 1 takes about 14 minutes

Lessons

3-6

Levels

4 free

Intro

11 paid

Advanced

Who this is for

People new to Claude Code

You know Claude chat, but repo work still feels like a separate world.

Repo owners

You want AI help without letting it touch auth, billing, data, or secrets by accident.

Teams standardizing AI work

You need shared instructions, review rules, and proof so sessions do not drift by person.

Power users

You are ready for hooks, MCP, skills, and subagents, but only when the setup earns its place.

Creative operators

You need code-agent help for sites or tools, but you still want visible proof before acceptance.

Automation builders

You want headless runs or SDK paths after the manual process has already worked.

4 Modules | 15 Lessons | Build in order

Module 1

4 lessons

Level 3 | 4 free lessons

Start safely

Use Claude Code only when the repo needs action, and undo without panic when it drifts.

Proof:You can classify a repo task, choose the surface, write a safe first-session brief, and rewind a session back to a known-good point.

New Claude Code usersRepo ownersTool-switchers

Start Module 1

Module 2

4 lessons

Level 4 | 4 paid lessons

Set repo context

Put instructions, settings, and task facts in the right place, then sort the task itself.

This module covers CLAUDE.md, the .claude folder, local settings, shared rules, clean session packets, and the recipe-versus-chef call: whether a task needs fixed steps or an open-ended goal.

Proof:You finish with repo instructions, setup placement rules, a clean context packet, and a real task sorted into a recipe or a chef.

Team leadsPower usersCourse authors

Start Module 2

Module 3

3 lessons

Level 5 | 3 paid lessons

Control action

Use permissions, review, and hooks before trust grows too loose.

This module teaches allow, ask, deny rules, diff review, proof checks, and hook decisions. The goal is controlled speed with clear stop points.

Proof:You finish with a permission map, review checklist, and one hook decision.

Repo maintainersReviewersAutomation owners

Start Module 3

Module 4

4 lessons

Level 6 | 4 paid lessons

Extend only when needed

Add MCP, skills, subagents, and headless runs only after proof, then verify by risk, not habit.

Proof:You finish with an MCP gap test, extension choice, automation readiness gate, and a verification contract for one real task.

Advanced operatorsTool buildersTeam system owners

Start Module 4