Skip to content

Vibe Coding Foundations

Claude Code types. You decide, and you know what good looks like.

AI coding agentsClaude CodecontextCLAUDE.mdspecsacceptance criteriaplan modereading codecode reviewdiffsgitcommitsbranchesundo

Lessons

  1. 38.1How AI coding agents work (and why "done" is not proof)What really happens when you ask Claude Code for a change, and why "done" still needs checking. 30 min
  2. 38.2Spec before code: decide what you want before the agent decides for youBefore the agent writes a line, write down what you actually want. Otherwise it decides for you. 30 min
  3. 38.3Plan before code: review the approach, not 1,400 linesAsk the agent how it will do the work before it touches a file. A short plan is far easier to check than a huge change. 30 min
  4. 38.4Reading code you didn't write (well enough to catch what matters)You don't have to write code to read it for what matters: what it sends, charges, saves or deletes, and when. 30 min
  5. 38.5Reviewing AI changes: read the diff, not the summaryThe agent's summary says what it meant to do. The list of changes shows what it did. Check the changes. 30 min
  6. 38.6Git as your safety net: save points, branches and the right undoSave often, try things on a side branch, and know which undo to use. Then no agent session can cost you more than a few minutes. 30 min

On the job: try it on your own project

Pick one small real change in your project. Before Claude Code writes anything, write a short spec with acceptance criteria, then ask it for a plan in plan mode and correct it. Let it work on a new branch, then read the diff yourself and accept, reject or ask for changes. Add one CLAUDE.md rule based on what you had to correct.

Can you do this now?

0/5. Anything you can't tick yet: reopen that lesson's practice ladder.