Cc Safe Setup

Features

• Warns on large file reads (>100KB)
• Limits unique file reads per session
• Estimates token budget and warns/blocks exceeding it
• Caps subagent spawns per session
• Provides hooks for various safety checks

Use Cases

• Protecting against accidental large file reads that consume excessive tokens
• Preventing uncontrolled spawning of subagents
• Monitoring and managing token usage to avoid unexpected costs
• Enhancing overall operational safety and stability of Claude Code

Non-Goals

• Replacing Claude Code's core functionality
• Directly managing user credentials or secrets
• Providing AI generation capabilities

Workflow

1. Install hooks via npm or plugin marketplace
2. Restart Claude Code to activate hooks
3. Hooks automatically monitor file reads, agent spawns, and token usage
4. Warnings or blocks are issued based on configurable thresholds
5. User adjusts behavior or CLAUDE.md based on hook feedback

Practices

• Operational safety
• Token management
• LLM agent security

Prerequisites

• jq installed on the system