Hook Selector

Answer 5 questions. Get your personalized Claude Code safety setup.
Based on 800+ hours of autonomous operation.

Question 1 of 5

How do you primarily use Claude Code?

Interactive — I'm always watching

I type commands and review each response before proceeding

Semi-autonomous — I check periodically

I start tasks and check back every 30-60 minutes

Fully autonomous — It runs while I sleep

I set tasks and let Claude Code work for hours unattended

Question 2 of 5

What's your biggest concern?

File deletion / data loss

I'm worried about rm -rf, accidental file overwrites, or git disasters

Token consumption / cost

My quota burns too fast, or costs spike unexpectedly

Secret leaks / security

I'm worried about .env exposure, credential leaks, or unsafe operations

Code quality / behavior

Claude Code ignores my instructions or writes bad code

Fabrication / fake results

Claude claims tests passed without running them, or fabricates warnings

Question 3 of 5

What's your technical comfort level?

Beginner — I don't write code

I need copy-paste solutions that just work

Intermediate — I can edit scripts

I can modify bash scripts and JSON configuration

Advanced — I build custom solutions

I write my own hooks and integrate with external tools

Question 4 of 5

Do you use subagents or MCP servers?

No — Just the main Claude Code

I use Claude Code in a single terminal session

Yes — Subagents (Agent tool)

I spawn subagents for parallel tasks

Yes — MCP servers

I use external tools via MCP protocol

Both — Subagents + MCP

I use the full ecosystem

Question 5 of 5

Have you experienced any incidents?

No incidents yet

Everything has been smooth so far

Minor issues — annoying but no damage

Infinite loops, wasted tokens, or ignored instructions

Major incident — real damage occurred

File deletion, secret exposure, or significant token waste