dcg (destructive_command_guard): Preventing Destructive AI Actions
The dcg tool acts as a protective layer that intercepts dangerous commands before they are executed by AI agents.
This prevents accidental data loss or disk formatting caused by autonomous tools like GPT-5 enough to protect files on devices such as MacBooks.
Key Features
- Intercepts potentially wayward commands before execution preventing file deletion and disk formatting
- Monitoring support for Claude Code, Codex, Gemini, Copilot, and Cursor
How to Use
- Install the dcg package via GitHub.
- Use it while working with AI coding assistants or agents who require command access.
Verdict: A vital safety net for developers using any form of high-level AI automation으로 protection against unintended system changes.
Note/Alternative Tooling: If you use Claude Code, please note there is also an internal /checkup command which cleans up unused skills, MCP servers, plugins, updates the software version, disables slow hooks, and optimizes project structure without risking damage through confirmation prompts.
! DYOR (Do Your Own Research)