C2. Claude for IT work

Outcome: you can use Claude Code to write and fix lab scripts, and you review every change before you keep it.

Before you start: S0 (command line on Windows and Linux, and Git) and a paid Claude plan or Anthropic Console account.

Skills: installing Claude Code, starting it in a project folder, CLAUDE.md project instructions, asking for a plan before any change, reading diffs, approving or refusing commands, Git commits as checkpoints, testing what it wrote, keeping secrets out of the repo.

Courses: Claude Code 101, Claude Code in action.

Labs

1. Install Claude Code on the lab host or the Linux VM, and start it in a new Git repo for your lab scripts.
2. Write a CLAUDE.md for the repo: what the lab is, the VM names from LAB-ENVIRONMENT.md, "lab only, never production", and how to test a script.
3. Ask for a PowerShell script that reports free disk space and stopped services on the lab servers. Ask for the plan first and read it, then let it write. Run the script after a snapshot. Commit.
4. Do the same in Bash for the Linux VM. Break one thing on purpose and ask Claude Code to find the fault from the error message.
5. Say no: when Claude Code proposes something you don't understand or wouldn't run yourself (a recursive delete, a change outside the repo), refuse it and ask for a safer way. Write down what happened.
6. Review log: for every change Claude Code made, note what you checked and what you fixed by hand.

Prove it: the scripts repo with its CLAUDE.md, the commit history, test output and the review log. A partner runs your scripts in their own lab from the README alone.

Certs: none.