Week 2 Overview

Last week you built your own project from scratch. This week is the opposite — and much closer to real work: entering a codebase you didn't write, making a change, and breaking nothing.

This is where beginners get hurt with an agent. On your own project, a bad change costs you a redo. In an existing project, the agent can quietly rewrite shared logic, duplicate code it never read, or "improve" things nobody asked it to touch. The skills of this week — directing the agent's reading, working in small reviewable steps, and setting explicit boundaries — are what make agent work safe on real codebases.

Learning goals

Understand how the agent explores an unfamiliar codebase — and the two practical consequences (it can duplicate what it hasn't read; precise direction improves results).
Use plan mode to explore and propose changes without touching a single file.
Work in a separate branch with small, reviewable diffs.
Isolate a feature in its own files and set explicit boundaries in prompts.
Treat any change to shared code as a mandatory manual review.

Resources

Official (source of truth):

Common workflows — exploring codebases, plan mode, everyday recipes: https://code.claude.com/docs/en/common-workflows
Claude Code 101 (course): https://anthropic.skilljar.com/claude-code-101

The HackYourFuture curriculum is licensed under CC BY-NC-SA 4.0 *https://hackyourfuture.net/*

CC BY-NC-SA 4.0 Icons

Built with ❤️ by the HackYourFuture community · Thank you, contributors

Found a mistake or have a suggestion? Let us know in the feedback form.