Skip to content
bits
Sign inStart free

Work with coding agents

A coding agent is Claude working in your repo: it reads files, edits them, runs your commands and checks its own work. This course shows you how to work with one well. You'll write the briefs, context files, skills and subagent definitions yourself, and Claude coaches each one. Everything is shown in Claude Code, the agent the build-along courses use next.

  • Beginner
  • 7 lessons
  • 53 min
  • text
Start the course →

Lesson 1 · A loop in your repo · 7 min

Syllabus

M1: The agent and you

  1. 1. A loop in your repo — not startedPrompt exercise7 min · +50 neurons
  2. 2. The file it reads every time — not startedPrompt exercise7 min · +50 neurons
  3. 3. Plan, build, review — not startedPrompt exercise8 min · +60 neurons

M2: Give it more to work with

  1. 4. Teach it a skill — not startedPrompt exercise8 min · +60 neurons
  2. 5. Give it new tools — not startedPrompt exercise7 min · +50 neurons

M3: Bigger jobs

  1. 6. Subagents and hooks — not startedPrompt exercise8 min · +60 neurons
  2. 7. Checkpoint: brief an agent — not startedPrompt exerciseProject8 min · +80 neurons

On finishing

Agent wrangler badgePlus 410 neurons and a certificate

Neurons are the points you earn for each lesson you finish.

Your progress stays in this browser. Sign in to keep it on every device.

What you'll learn

  1. Explain what a coding agent does in your repo: a loop that reads, edits, runs and checks.
  2. Write a short CLAUDE.md with what the agent can't work out for itself.
  3. Brief an agent so it plans first, keeps the change small and proves it worked.
  4. Add skills and MCP servers, and know when a subagent or a hook is the right tool.

How you'll learn

7
lessons, every one a chat with Claude
7
prompt exercises
1
checkpoint project

Source

This course is 8 markdown files on GitHub. Spot a typo? Fix it in one pull request.

content/courses/work-with-coding-agents/ ↗

Written by The bits team · updated

Code verified