Skip to content
Module
Module 6 of 6Lesson 1 of 3~18 min

Put an AI coding agent to work in a team repository

An AI coding agent writes fast, but without a framework it can bypass the team's habits and safeguards. This lesson helps you bring it into the repository's usual workflow, with written rules, proof of verification, secrets kept safe and a human review before every merge.

Lesson objective

By the end of this lesson, you will be able to make a coding agent such as Claude Code work inside the team's workflow: on a dedicated branch, with rules written in CLAUDE.md, proof of verification, protected secrets and a human review of every diff before merging.

Topics covered

  • AI coding agent
  • Claude Code
  • CLAUDE.md
  • code review
  • secrets management

Where it fits

Coding agent, releases and end-to-end delivery

How do I bring an AI agent into this workflow, mark a release and ship my first improvement?

Lessons in this module

  1. Put an AI coding agent to work in a team repository (this lesson)
  2. Tags, versions and release notes
  3. Final workshop: ship an improvement end to end

What you will learn in this module

This lesson is part of the course Git for PMs: ship as a team without putting production at risk

  • Work on a short-lived branch with everyday Git commands (or their GitHub equivalents), small commits named by convention, and stay in sync with main.
  • Open a small, well-described pull request, get the right review, pass the required checks, validate the preview and choose the merge method.
  • Choose the right way to undo a change (restore, revert, reset, deployment rollback) depending on whether it is local, pushed, merged or already in production.
  • Keep an AI coding agent under control in a team repository (dedicated branch, written rules, protected secrets, proof of verification, human review before merge).
  • Tie merged changes to a release (tag, semantic versioning, changelog) and write release notes that help users.