Advanced
Claude Code: Safety
Master the permission system, sandboxing, auto-mode classifier, credential protection, and enterprise policy — everything you need to run Claude Code autonomously without risk.
01
The Agentic Threat Model
›
02
Permission Modes: Six Settings, One Right Choice
›
03
Permission Rules: Allow, Deny, and Ask
›
04
Protected Paths and Workspace Trust
›
05
Sandboxing: OS-Level Enforcement
›
06
Credential Protection
›
07
Auto Mode: How the Classifier Works
›
08
Configuring Auto Mode for Organizations
›
09
Prompt Injection: Threats and Defenses
›
10
Managed Settings: Enterprise Policy
›
11
Hooks as Security Gates
›
12
Production Guards: CI/CD and Unattended Runs
›