Skip to main content

Overview

Dangerous Skip Permissions mode bypasses the permission system entirely, allowing tools to execute immediately without user approval. This is designed for automated workflows, testing, and background tasks where interactive prompts are not possible.
This mode is called “dangerous” for a reason. Only use it in controlled environments with proper security hooks.

How It Works

When enabled:
  1. Hooks still run — security hooks can still block operations
  2. Permission system bypassed — global/local/session rules are ignored
  3. No user prompts — tools execute immediately (if hooks allow)

Enabling Dangerous Mode

At Agent Creation

At Runtime

When to Use

Appropriate:
  • CI/CD pipelines
  • Automated testing
  • Background tasks with no UI
  • Trusted server-side environments (with security hooks)
Never use for:
  • Interactive user-facing applications
  • Agents processing untrusted input
  • Production environments without security hooks
  • Multi-tenant applications

Security with Hooks

Always use security hooks with dangerous mode. Hooks are your safety net when permissions are disabled.

Conditional Configuration

Next Steps

Three-Tier System

Normal permission system

Hooks

Security hooks