Dicklesworthstone/destructive_command_guard

The Destructive Command Guard (dcg) is for blocking dangerous git and shell commands from being executed by agents.

What it solves

AI coding agents (such as Claude Code, Cursor, and GitHub Copilot) can occasionally execute catastrophic commands—like rm -rf ./src or DROP TABLE users—that destroy uncommitted work or production data in seconds. dcg provides a safety layer to intercept and block these destructive commands before they are executed.

How it works

It acts as a high-performance hook that intercepts commands sent by AI agents. It uses SIMD-accelerated filtering and a modular "pack" system to scan commands against known destructive patterns. It can detect dangerous operations even when embedded in heredocs or inline scripts (e.g., python -c "os.remove(...)"). If a command is flagged, dcg blocks the execution and displays a human-readable denial panel with the reason and safer alternatives on stderr, while keeping machine-readable output on stdout for the agent.

Who it’s for

Developers using AI coding agents or CLI-based AI tools who want to prevent accidental data loss or system destruction caused by autonomous agent actions.

Highlights

  • Broad Agent Support: Integrates with Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Grok, and many others.
  • Modular Security Packs: Over 50 specialized packs covering databases (PostgreSQL, MongoDB), cloud providers (AWS, GCP, Azure), containers (Docker, Kubernetes), and infrastructure (Terraform, Pulumi).
  • Sub-Millisecond Latency: Designed for minimal overhead so the protection doesn't slow down the development workflow.
  • Smart Context Detection: Distinguishes between data (e.g., grep "rm -rf") and actual execution (e.g., rm -rf /).
  • Flexible Bypass: Offers multiple ways to allow commands, including environment variables (DCG_BYPASS=1), one-time allow codes, or permanent allowlists.

Related

  • Project
    manuelschipper/nah

    A deterministic safety guard for AI coding agents that blocks catastrophic tool calls like filesystem deletion and secret theft without using an LLM.

  • Project
    kenryu42/cc-safety-net

    A security layer for AI coding CLIs that blocks destructive commands and access to sensitive secrets before they are executed by an agent.

  • Project
    hashgraph-online/hol-guard

    An open-source security layer that provides antivirus-style runtime protection for AI agents, blocking risky actions like secret leaks and unsafe commands.

  • Project
    hardenrun/aif

    A local safety system for coding agents that prevents the execution of dangerous commands and blocks secret leaks through pre-execution blocking.

  • Project
    colbymchenry/codegraph

    CodeGraph is a Rust‑based local tool that builds a full‑code knowledge graph for any project (supports dozens of languages) and integrates it with AI coding assistants (Claude Code, Cursor, Copilot, Gemini, etc.). The graph lets agents retrieve exact symbols and call‑paths in a single request, cutting tool calls, file reads, token usage and cost dramatically. Install via a one‑line script, run `codegraph init` per repo, and the CLI auto‑updates the graph as you edit. A browser UI (`codegraph ui`) visualises the graph, and the CLI can uninstall itself cleanly.