← 返回 Skills 市场
265
总下载
0
收藏
1
当前安装
1
版本数
在 OpenClaw 中安装
/install self-guardian
功能描述
Lightweight self-monitoring and self-constraining rules to prevent accidental file deletions, dangerous commands, and risky operations. Activates automatical...
安全使用建议
This skill is instruction-only and appears coherent for the stated purpose (preventing accidental destructive file/shell operations). Before installing: 1) Review the full SKILL.md (the prompt showed it truncated) to ensure there are no hidden steps that access network endpoints, read unrelated files, or request persistence. 2) Understand operational impact: it will pause or require confirmation for many destructive commands and sensitive files — expect more prompts when doing recovery or blanket operations. 3) Keep it user-invocable (do not set always:true) until you have tested it in a safe repository to confirm it doesn't block legitimate workflows. If you want extra assurance, run a manual audit of the entire SKILL.md text (especially the truncated portion) and verify there are no instructions to fetch code or exfiltrate data.
功能分析
Type: OpenClaw Skill
Name: self-guardian
Version: 0.1.0
The 'self-guardian' skill is a safety-oriented utility designed to prevent accidental data loss or dangerous system operations by the AI agent. It implements a pre-action checklist, defines file protection tiers (specifically protecting sensitive items like .env files, SSH keys, and agent data directories), and establishes a blocklist for destructive commands (e.g., rm -rf, sudo, git push -f) that requires explicit user confirmation. The skill contains no malicious logic or exfiltration attempts and serves as a set of guardrails for the agent's behavior.
能力评估
Purpose & Capability
Name/description match the skill contents: the SKILL.md defines non-negotiable pre-action checks, file protection tiers, command blocklists and throttles. The skill requests no binaries, env vars, or installs — all of which is proportionate for a policy/enforcement skill.
Instruction Scope
The instructions require the agent to perform internal checks before any file/write/delete or shell command and to ask the user when in doubt; this stays within the stated safety purpose. Note: the provided SKILL.md in the prompt includes a truncation marker, so I could not review the very end of the file; if the truncated portion adds behaviors (network calls, reads of unrelated files, or persistence instructions), that could change the assessment.
Install Mechanism
No install spec and no code files — instruction-only. This minimizes risk because nothing is downloaded or written to disk by the skill itself.
Credentials
The skill declares no required environment variables, credentials, or config paths. It sensibly enumerates sensitive files and directories to avoid touching (SSH keys, .env, DB files, agent data dirs), which is appropriate for a safety policy.
Persistence & Privilege
always: false and no requests to modify other skills or global agent config. The skill can be invoked autonomously by the agent (default), which is normal for skills; nothing here grants elevated or persistent system privileges.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install self-guardian - 安装完成后，直接呼叫该 Skill 的名称或使用
/self-guardian触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v0.1.0
Initial release: Pre-Action Checklist, file protection tiers, destructive command blocklist, batch operation throttle, model-awareness heuristic, recovery breadcrumbs
元数据
常见问题
Self-Guardian 是什么？
Lightweight self-monitoring and self-constraining rules to prevent accidental file deletions, dangerous commands, and risky operations. Activates automatical... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 265 次。
如何安装 Self-Guardian？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install self-guardian」即可一键安装，无需额外配置。
Self-Guardian 是免费的吗？
是的，Self-Guardian 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Self-Guardian 支持哪些平台？
Self-Guardian 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Self-Guardian？
由 vaderyang（@vaderyang）开发并维护，当前版本 v0.1.0。
推荐 Skills