Things to Watch Out For

Security Considerations

What Claude Code CAN do:

• Read any file in current directory

• Write/modify files

• Execute terminal commands

• Install packages (if you confirm)

What it CAN’T do (without you noticing):

• Access files outside your project folder (unless you cd there)

• Run commands without showing you

• Install system-wide software (needs sudo password)

Golden rule:

• Always review code before running

• Use Git so you can undo

• Start in dedicated project folders

Privacy Mindset

Remember:

• ⚠️ Check your settings: claude.ai/settings/data-privacy-controls

• Consumer accounts: allow data use → retention up to 5 years; disallow → 30 days

• All requests are processed by Anthropic’s servers (encrypted in transit)

For sensitive projects:

• Consider: Running on separate machine

• Or: Use tools with privacy modes (verify their policies)

• Or: Local-only AI models (more technical, less capable)

Cost Management

Plans and limits change:

• Usage limits and context windows vary by plan

• Check your account settings for current limits and upgrade options

Watch out for:

• Long-running tasks can use more of your quota

• If you hit limits, split tasks or upgrade your plan

The “Magic Black Box” Problem

Don’t fall into trap of:

• Accepting all code without reading

• Not understanding basic structure

• Becoming dependent without fundamentals

Better approach:

• Ask Claude to explain major sections

• Read the comments it writes

• Modify small things yourself to learn

---