⏳ This skill is pending AI review.

Scores will appear once the review pipeline completes.

version unknown

hook-guard

@wavmson⭐ 0 stars

钩子守卫。为 Agent 操作添加安全防护层——文件修改前自动备份、危险命令执行前拦截确认、敏感操作自动告警通知用户。触发词：安全检查、hook guard、守卫、备份保护、安全模式、操作审计。也可在 AGENTS.md 中配置为始终生效。

Choose how to use this skill

You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.

Adopt or update this skill from https://proskills.md/skills/other/wavmson-openclaw-skill-hook-guard. First identify the installation path supported by this platform, detect whether it is already installed, compare the installed version with the latest stable release, show me what will change, and only then install or update it. If no native installer is available, use the complete package or source fallback described on the page.

1. Native installer

This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.

Do not guess an installer command or replace an existing version without reviewing the diff.

2. Complete package recommended

Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.

No complete ProSkills package is published for this listing yet.

3. Prompt-only

Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.

Need only the instruction file?

Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.

No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.

—/10

// RATINGS

⭐GitHub Stars

⭐ 0 on GitHubGitHub ↗

New / niche

🟢ProSkills Score

—

📍

Not yet listed on ClawHub or SkillsMP

// README

🛡️ Hook Guard — 钩子守卫 Skill

给你的 AI Agent 装上安全气囊。危险操作拦截确认，重要文件自动备份，所有行为可审计追溯。

中文说明

这是什么？

AI Agent 拥有强大权限：读写文件、执行命令、发消息、改配置。但能力越强越容易出事：

😱 Agent 误删了重要文件
💥 一个 rm -rf 清空了项目目录
📤 没确认就把消息发到了老板群
⚙️ 改错配置导致服务挂了

Hook Guard 是 Agent 的安全防护层。三级钩子系统覆盖所有操作。

安装

通过 ClawHub 安装（推荐）：

clawhub install hook-guard

从 GitHub 克隆：

git clone https://github.com/wavmson/openclaw-skill-hook-guard.git \
  ~/.openclaw/skills/hook-guard

安装后重启 Gateway：

openclaw gateway restart

三级防护体系

级别	名称	触发场景	处理方式
🔴 Red	危险拦截	删除文件、sudo、发布到公开平台	暂停执行，等待用户确认
🟡 Yellow	自动备份	修改配置、覆盖文件、改 cron	备份后执行，通知用户
🟢 Green	静默日志	读取文件、搜索、查看信息	只记录日志，不打扰

Red Hook 示例

Agent 准备执行危险操作时：

🔴 危险操作拦截
━━━━━━━━━━━━

操作：rm -rf /home/wavm/important-project/
影响：将永久删除 342 个文件
风险：不可逆操作

建议：
1. 改用 trash（可恢复）
2. 先备份到 /tmp/

等待确认（回复"继续"执行，"取消"放弃）

Yellow Hook 示例

修改配置文件时自动备份：

🟡 自动备份
文件：openclaw.json
备份：.hook-guard/backups/2026-04-02/openclaw.json.143022.bak
操作：修改 maxTokens 配置

守卫报告

说"守卫报告"或"guard report"查看统计：

🛡️ Hook Guard 报告（最近 24 小时）

🔴 拦截：2 次
🟡 备份：5 次（2.3 MB）
🟢 日志：47 条操作记录

最近拦截：
  - rm -rf /tmp/old-project/ → 改用 trash
  - sudo systemctl restart nginx → 确认后执行

设计原则

原则	说明
🔒 不可绕过	Red Hook 危险操作必须单次确认，即使用户说过"别问了"
💾 自动备份	Yellow 操作自动保存原始状态，7 天自动清理
📝 全程可追溯	所有操作记录审计日志，30 天保留
⚡ 不影响性能	Green 日志异步写入，零延迟
🎚️ 可调级别	用户可以升级或降级某些操作的防护等级

哪些操作会被拦截？

🔴 Red（必须确认）：

删除文件（rm、rm -rf）
含 sudo 的命令
停止系统服务
清理 Docker 容器/镜像
发送消息到新聊天
发布到公开平台
修改 DNS/域名
操作生产环境

🟡 Yellow（自动备份）：

修改配置文件
覆盖写入工作区文件
修改 cron 定时任务
更新数据库记录

🟢 Green（只记日志）：

读取文件
搜索操作
查看日历/任务
网络搜索

与记忆保护链搭配

Skill	职责
Smart Compact	压缩前抢救信息
Session Resume	断线后恢复进度
Memory-Dream	定期整合长期记忆
Swarm Coord	多 Agent 并行协作
Hook Guard	操作安全防护

五个 Skill 形成完整的 Agent 安全运维体系：

记忆不丢（Smart Compact + Memory-Dream）
任务不断（Session Resume）
效率不低（Swarm Coord）
操作不炸（Hook Guard）

常见问题

Q: 会不会每次都弹确认很烦？ A: 不会。只有 Red 级操作需要确认，日常读写都是 Green 级静默记录。

Q: 备份会不会占很多空间？ A: 只备份被修改的单个文件，7 天自动清理，通常只占几 MB。

Q: 子 Agent 也受保护吗？ A: 是的。Swarm Coord 启动的子 Agent 同样遵守 Hook Guard 规则。

Q: 能自定义哪些操作需要确认吗？ A: 可以。在 AGENTS.md 中配置自定义规则，升级或降级操作等级。

English

The Problem

AI agents have powerful permissions — file read/write, command execution, message sending, config changes. With great power comes great risk:

Accidentally deleting important files
Running destructive commands without confirmation
Sending messages to wrong recipients
Breaking configurations

The Solution

Hook Guard adds a safety layer with three protection levels:

Level	Name	Trigger	Action
🔴 Red	Dangerous	Delete, sudo, publish	Pause and ask user
🟡 Yellow	Important	Config edits, overwrites	Auto-backup, then proceed
🟢 Green	Routine	Read, search, browse	Silent audit log

Install

clawhub install hook-guard

Or clone from GitHub:

git clone https://github.com/wavmson/openclaw-skill-hook-guard.git \
  ~/.openclaw/skills/hook-guard

How It Works

Before every operation:

Classify: Red, Yellow, or Green?
Red → Generate interception report → Wait for user confirmation
Yellow → Auto-backup original state → Execute → Notify user
Green → Log silently → Execute normally

Design Principles

Principle	Description
🔒 Non-bypassable	Red operations always need per-operation confirmation
💾 Auto-backup	Yellow operations save original state (7-day retention)
📝 Full audit trail	All operations logged (30-day retention)
⚡ Zero overhead	Green logs are async, no performance impact
🎚️ Customizable	Users can adjust protection levels per operation

The 5-Skill Agent Safety Stack

Skill	Protects
Smart Compact	Conversation details
Session Resume	Task progress
Memory-Dream	Long-term memory
Swarm Coord	Team efficiency
Hook Guard	Operation safety

FAQ

Q: Will it interrupt me constantly? A: No. Only Red-level operations (deletions, sudo, publishing) need confirmation. Daily work is Green-level (silent).

Q: How much space do backups use? A: Only modified files are backed up. 7-day auto-cleanup keeps it under a few MB.

Q: Does it protect sub-agents too? A: Yes. Agents spawned by Swarm Coord follow the same Hook Guard rules.

Requirements

OpenClaw with Skill support

License

MIT

Author

@wavmson

// HOW IT'S BUILT

KEY FILES

SKILL.mdREADME.md

// REPO STATS

0 stars

// SHARE

Share on X Share on LinkedIn