SafeAI-Lab-X/ClawKeeper

Registry 信息

项目简介

ClawKeeper 是一个为 OpenClaw 代理提供全面安全保护的框架，通过技能、插件和监视器实现分层防护。它常被称为“OpenClaw 的诺顿”，帮助开发者和用户防范恶意或意外的代理行为。

要点

• 通过技能、插件和监视器实现分层安全
• 实时监控代理行为
• 基于插件的可扩展安全架构
• 专为 OpenClaw 代理设计
• 社区驱动的安全规则与策略

使用场景

• 在生产环境中保护 OpenClaw 代理
• 防止代理未经授权访问数据
• 在多代理系统中强制执行安全策略
• 审计和记录代理行为以满足合规要求
• 针对特定威胁开发自定义安全插件

README 摘要

# 🦞🛡️ ClawKeeper: Comprehensive Safety Protection for OpenClaw Agents Through Skills, Plugins, and Watchers <h1 align="center"><i>(aka The Norton for OpenClaw)</i></h1> <p align="center"> <img src="./fig/logo.png" alt="OpenClaw" width="700"> </p> <p align="center"> <strong>SAFETY EXFOLIATE! SAFETY EXFOLIATE!</strong> </p> <p align="center"> <a href="https://github.com/openclaw/openclaw"> <img src="https://img.shields.io/badge/OpenClaw-Compatible-blue.svg" alt="OpenClaw"> </a> <a href="https://opensource.org/licenses/MIT"> <img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"> </a> </p> **ClawKeeper** is a _comprehensive real-time security framework_ designed for autonomous agent systems such as **OpenClaw**. It provides unified protection through three complementary approaches: **skill-based** safeguards at the instruction level, **plugin-based** enforcement at the runtime level, and a **watcher-based** independent monitoring agent for external oversight. # 🔎 Overview **ClawKeeper** provides protection mechanisms across three complementary architectural layers: - **Skill-based Protection** operates at the instruction level, injecting structured security policies directly into the agent context to enforce environment-specific constraints and cross-platform boundaries. - **Plugin-based Protection** serves as an internal runtime enforcer, providing configuration hardening, proactive threat detection, and continuous behavioral monitoring throughout the execution pipeline. - **Watcher-based Protection** introduces a novel, decoupled system-level security middleware that continuously verifies agent state evolution. It enables real-time execution intervention without coupling to the agent's internal logic, supporti

话题