← 返回 Skills 市场
AgentGuard
作者
manas-io-ai
· GitHub ↗
· v1.0.0
4272
总下载
3
收藏
34
当前安装
1
版本数
在 OpenClaw 中安装
/install agentguard
功能描述
Monitors agent file access, API calls, and communications to detect suspicious behavior, log events, and generate actionable security reports.
安全使用建议
Install only if you intentionally want an agent activity monitor. Before starting it, limit watched directories, keep alerting/reporting console-only unless you explicitly want data sent to Telegram/Discord/webhooks, review what gets written under ~/.agentguard, shorten retention if needed, and verify the publisher/source and dependencies before running the Python scripts.
功能分析
Type: OpenClaw Skill
Name: agentguard
Version: 1.0.0
The AgentGuard skill bundle is designed for security monitoring of AI agent operations. Its core functionality involves detecting suspicious file access (e.g., sensitive files like `.env`, `.ssh`), API calls (e.g., to untrusted domains, credential exposure in URLs), and potential data exfiltration patterns. The Python scripts implement these detection and logging capabilities, explicitly sanitizing and hashing sensitive data before logging to prevent exposure. The `SKILL.md` and `README.md` accurately describe these security features and do not contain any prompt injection attempts or instructions for malicious behavior. All observed behaviors are aligned with the stated purpose of enhancing agent security.
能力评估
Purpose & Capability
The file, API, communication logging, alerting, and reporting features fit a security-monitoring skill, but the artifacts also describe Telegram/report channels, inter-skill threat sharing, and optional blocking while claiming no external data transmission.
Instruction Scope
The skill discloses broad monitoring, persistent logs, and reports, but does not clearly define consent, recipient boundaries, minimization rules, or approval controls for external channels and inter-skill sharing; the summary command also starts monitoring briefly.
Install Mechanism
The README gives a normal ClawHub install command and the package has executable Python scripts, but metadata shows no parsed install spec or source homepage. I found no hidden install-time execution or obfuscated setup behavior.
Credentials
Default watched paths are agent-related directories, but the monitor records file paths, domains, URLs, communication destinations, sensitive-file classifications, and timing metadata that can reveal private workspace and credential-handling activity.
Persistence & Privilege
The skill creates ~/.agentguard storage, keeps logs, baselines, alerts, and reports, supports daemon-style monitoring, and advertises automatic reports; this is purpose-aligned but high-impact enough to require clearer retention and shutdown controls.
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 agentguard - 安装完成后，直接呼叫该 Skill 的名称或使用
/agentguard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
v1.0.0
Initial release of AgentGuard – security monitoring skill.
- Monitors file access and API calls for suspicious activity, including credential access and exfiltration patterns.
- Logs agent communications (HTTP requests, emails, message platforms) with audit trails.
- Detects behavioral anomalies using ML-lite pattern analysis.
- Generates security reports with alerts, activity summaries, and recommendations.
- Features configurable alert channels, data retention, sensitivity levels, and privacy-focused data handling.
- Integrates with Clawdbot and can share/block operations with other skills.
元数据
常见问题
AgentGuard 是什么？
Monitors agent file access, API calls, and communications to detect suspicious behavior, log events, and generate actionable security reports. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 4272 次。
如何安装 AgentGuard？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install agentguard」即可一键安装，无需额外配置。
AgentGuard 是免费的吗？
是的，AgentGuard 完全免费（开源免费），可自由下载、安装和使用。
AgentGuard 支持哪些平台？
AgentGuard 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 AgentGuard？
由 manas-io-ai（@manas-io-ai）开发并维护，当前版本 v1.0.0。
推荐 Skills