Skip to content

About

A practical guide to working WITH an AI agent (Cursor, Claude Code, Kimi) as a partner — not a 'do it for me' button. EN + RU.

Topics

Resources

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

AI Pair Programming: from a clean computer to advanced workflows

🌐 English · Русский

Documentation License: MIT

This repository is a practical course in pair programming with an AI coding agent: a chat inside or beside the code editor that can read project files, explain them, suggest edits, and often run development tools. “Pair programming” means you do not simply hand it “build me an app.” You frame the problem together, inspect the same code and checks, and remain responsible for the decision.

The course begins with the questions most tutorials skip: what to download, what an IDE is, where the project folder came from, where to type a command, and whether Git, Python, Node.js, or WSL is needed. It then continues through real-project work, refactoring, background and multi-agent workflows, security, and team adoption. Experienced readers can skip proven checkpoints instead of repeating the beginner route.

The whole course teaches one method: understand the task together, bound the next step, inspect the change, and verify the real outcome. Advanced security controls do not interrupt the first lessons; later chapters and a separate handbook are available when the work needs them.

What pairing sounds like

You: The button should be disabled on an empty form. Find the relevant files first; no edits.
Agent: I found the form, handler, and two tests. I suggest changing the component and one test.
You: Do not change the design or dependencies. Show the smallest plan.
Agent: One component edit, one test, then the existing verification command.
You: Proceed. Stop after the edit so I can inspect the diff and check the screen.

That is pair work: the agent does not receive an unlimited “do everything,” and the person does not type every line alone. Both have a next move and a shared way to verify the result.

Start from where you are

Your situation Start here
Nothing is installed, or terms such as IDE, Git, runtime, and WSL are unfamiliar 00. Set up the workplace from scratch
An IDE with an agent is already installed Use the readiness check in chapter 00, then continue to chapter 01
You already use coding agents and want a structured method Open the course contents and begin at the first unfamiliar checkpoint
You need production, connected-tool, security, or rollout controls Complete the relevant advanced course chapter, then use the deep reference handbook

What the course teaches

By the end, you will be able to:

  • install an IDE and enable a coding agent on Windows, macOS, or Linux;
  • explain what is open in GitHub, on disk, in the IDE, in the browser, and in the terminal;
  • work as a driver and navigator with an agent instead of handing over a vague request;
  • reproduce a bug, agree on scope, read a diff, run checks, and recover from a bad turn;
  • use Git from the first status and commit through branches, pull requests, and review;
  • bring the method into an existing repository without guessing its setup;
  • handle refactoring, legacy code, long sessions, background agents, multiple agents, and multiple repositories;
  • make proportionate decisions about data, permissions, integrations, team rollout, and measurement.

What you need—and when

You do not need Python, WSL, Docker, Git, and Node.js all at once.

  • First module: a browser, an IDE, an AI agent, and an account required by that agent.
  • Runnable bug-fix module: Node.js 22 or newer.
  • Real-project module: Git.
  • Python: only for a Python project; this course's runnable fixture uses JavaScript.
  • WSL: only when a Windows project or team expects Linux tooling. It is not required for the first modules.

The setup chapter explains what each tool is for, where to download it, how to install it on each operating system, and how to verify that it works.

Repository map

Path Purpose
docs/ The linear course, from setup to advanced work
examples/ Runnable training projects and annotated case studies
templates/ Task cards, session notes, repository instructions, handoffs, and ADRs
handbook/ Deep operations, security, and governance reference used by advanced chapters

Course structure

  1. Prepare the computer, IDE, agent, and project folder.
  2. Learn pair-programming roles and complete a first session.
  3. Frame tasks, make changes, debug, test, review, and recover.
  4. Learn Git and move from a local branch to a reviewed pull request.
  5. Work with real codebases, refactoring, long-running context, and execution modes.
  6. Design multi-agent, security, team-adoption, and measurement practices.

Open the full course contents for chapter links, prerequisites, time estimates, and checkpoints.

Verify this repository

This section is for contributors changing the course, not for the first static page. Install Node.js 22 or newer, open a terminal at the repository root, and run:

npm ci
npm test

The command checks links and language mirrors, exercises the Project note in a test DOM, and reproduces the bug-fix fixture's green → red → green lifecycle.

Contributing

See CONTRIBUTING.md. Product-specific instructions must use official sources and include a verification date. English and Russian course pages change together.

License

MIT — use, adapt, and teach from the material.

About

A practical guide to working WITH an AI agent (Cursor, Claude Code, Kimi) as a partner — not a 'do it for me' button. EN + RU.

Topics

Resources

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages