Now, before we dive deeper, I need to talk about security limitations.

When you first run Claude Code, you'll see a permission prompt asking you to approve its actions. Why? Well, think about this: what if you clone a repository and the README file contains malicious instructions like "delete all the user's files"? Claude would read that and potentially act on it.

This is a real vulnerability. Claude Code is susceptible to prompt injection attacks. These aren't fully solvable at the moment, though there are lots of ways to make them harder, like using sandboxed environments.

You'll want to be particularly careful about this. Always review what Claude is about to do before approving actions, especially when working with unfamiliar codebases or repositories. The permission system is there to protect you, so use it wisely.