AgentGuard is a robust security layer that monitors agent operations, detects suspicious activities, and generates actionable security reports in real-time.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agentguard
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agentguard using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is AgentGuard?
AgentGuard provides a comprehensive defense-in-depth framework for managing AI agent activities within your local or cloud environment. As one of the most critical Openclaw Skills for production-grade deployments, it monitors file access, outbound API calls, and external communications to ensure agents operate within safe boundaries. By leveraging ML-lite pattern analysis, it identifies behavioral deviations and potential exfiltration attempts before they become security breaches.
This skill is designed with the assumption that agents can be compromised or manipulated, providing an essential layer of visibility into every operation. It works by establishing a baseline of normal behavior and alerting administrators to any deviations, such as access to credential files or unusual API call frequencies, making it an indispensable part of any secure Openclaw Skills ecosystem.
AgentGuard Use Cases
- Monitoring unauthorized access to sensitive credential files like .env, keys, or secrets.
- Detecting outbound API calls to untrusted or malicious domains in real-time.
- Generating detailed audit trails for compliance and security forensics.
- Identifying volume spikes or unusual operation chains through automated anomaly detection.
- Preventing data exfiltration by blocking suspicious network calls following large file reads.
How AgentGuard Works
- Initialize the monitoring daemon to establish a baseline of normal agent behavior and resource usage.
- Track file system operations and API requests continuously across the configured watch directories.
- Analyze incoming activity streams using the anomaly detection engine to flag deviations based on volume, sequence, or destination.
- Trigger instant alerts through configured communication channels when high-severity threats are detected.
- Compile automated daily or weekly security reports summarizing all activity, alerts, and resource access patterns.
AgentGuard Setup
To get started with this security skill, follow these installation steps:
# Navigate to your skill directory and install dependencies
pip install agentguard
# Configure your monitoring preferences and trusted domains
nano ~/.agentguard/config/agentguard.yaml
# Start the full monitoring suite with default settings
agentguard start
# Verify the current security status and active monitors
agentguard status
AgentGuard Data Schema & Taxonomy
AgentGuard organizes its security data and logs within a structured directory for easy auditing:
| Directory | Contents |
|---|---|
~/.agentguard/logs/ |
Detailed logs for file access, API calls, and communications. |
~/.agentguard/baselines/ |
JSON-formatted behavior models used for anomaly detection. |
~/.agentguard/alerts/ |
Daily JSON files containing categorized security events. |
~/.agentguard/reports/ |
Markdown-formatted summaries of security posture and recommendations. |
AgentGuard Advanced Features
- ML-powered anomaly detection with adjustable sensitivity levels (low, medium, high).
- Automated domain whitelisting to prevent false positives for trusted API integrations.
- Encrypted log storage using AES to protect sensitive audit data from tampering.
- Multi-agent support through shared threat intelligence and integration hooks.
- Configurable data retention policies to automatically manage log growth and privacy.
SKILL.md
Loading
Related Openclaw Skills
ClawPlace Agent API
A specialized API bridge for AI agents to participate in collaborative pixel art battles and faction-based territory control on the ClawPlace canvas.
Birthday Reminder
A natural language-driven birthday management skill that tracks special dates, calculates upcoming ages, and provides automated reminders.
MAMPE Industrial Smart Asset Transformation
A specialized AI skill for mechatronics, sensor integration, and industrial asset optimization managed through Openclaw Skills.
ShadowStrike Security
A professional-grade penetration testing and security assessment platform featuring over 600 specialized tools for automated security audits.
LegalDoc AI
LegalDoc AI is a professional automation skill for contract analysis, legal research, and automated deadline management for law firms and legal teams.
Lead Automation Skill
An end-to-end automation engine for lead discovery, dynamic landing page creation, and personalized email outreach sequences.
Featured*
