Claude Code in Action

What the course covers

Claude Code in Action is a free Anthropic Academy course about running Claude Code past the quick task and trusting the result. Its curriculum has four parts and a quiz:

• Steer the work: scoping long sessions with plan mode, directing compaction so summaries keep what matters, using the rewind menu to course-correct, and choosing between hands-on steering and autonomous runs.
• Configure Claude: a lean CLAUDE.md that Claude actually follows, verification skills, permission modes, and hooks that enforce rules the model cannot skip.
• Automate repeat work: routines, headless mode for your own pipelines, and GitHub Actions and code review on pull requests.
• Verify and share: verifying unsupervised runs in proportion to how little you watched them, gating turns on real test results, and packaging a trusted setup as a plugin.

Who it is for

Anthropic describes the audience as developers who already use Claude Code for single prompts and want to move to longer, less supervised, team-wide workflows. If you have never used Claude Code, take Claude Code 101 first.

Is it worth your time?

Yes, and especially for its emphasis on verification. The course's central idea is that the less you watched a run, the more you must check its result, with tests and hooks rather than trust. That is the right instinct for anyone letting an agent work on real code, and it is easy to skip when a tool feels impressive.

What you can do after it

• Keep a long session on track with plan mode, targeted compaction and rewind.
• Cut a CLAUDE.md down to the lines that change behaviour.
• Pick the permission mode that fits a job, and enforce hard rules with hooks.
• Run Claude Code headless in a script or CI, and wire it into pull request review.
• Decide how much evidence you need before trusting an unsupervised run.

Prerequisites

Comfort with Git, a terminal and your project's test commands, plus access to Claude Code through a paid Claude plan or an API key. The course itself is free.

How it fits the Coding with AI tools roadmap

The Claude Code edition of the Sefism roadmap uses it in three places: going deeper on prompting and context, learning permission modes and hooks in the agents stage, and shipping a real project with CI in the final stage.

Common questions

Does it cover security? Partly, through permission modes and hooks. Pair it with the roadmap's security stage for secrets, dependencies and insecure patterns.