MoltGuard is a comprehensive security framework that protects AI agents from prompt injection, data exfiltration, and malicious command execution.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install superguard
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 superguard 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 MoltGuard Security Guard?
MoltGuard, developed by OpenGuardrails, serves as a critical security layer for your AI coding agent. By integrating directly into the Openclaw Skills ecosystem, it provides real-time monitoring and protection against a wide array of vulnerabilities that can compromise both the agent and the host system. It is designed to act as an automated security guard, filtering out harmful instructions hidden in web content or files.
This skill focuses on mitigating three primary risk surfaces: instruction-based risks like prompt injection, behavioral risks like unauthorized file deletion, and data risks such as PII exposure or secret leakage. By utilizing intent-action mismatch detection, it ensures that your agent does not deviate from its intended safe operations, providing peace of mind for developers using Openclaw Skills in production environments.
MoltGuard Security Guard Use Cases
- Safeguarding agents against indirect prompt injection from malicious third-party websites or emails.
- Preventing the accidental execution of destructive shell commands or unauthorized system modifications.
- Identifying and blocking the leakage of sensitive data, such as API keys and personally identifiable information (PII).
- Implementing enterprise-grade security policies across multiple AI agent instances.
How MoltGuard Security Guard Works
- The plugin intercepts all incoming instructions and outgoing actions within the agent environment.
- Commands and content are analyzed by the Core security engine to detect potential risks.
- The system evaluates the intent of the agent against the actual requested action to find mismatches.
- If a risk is identified, MoltGuard blocks the malicious activity and alerts the user.
- Status and quota usage are reported back through the Openclaw Skills command interface.
MoltGuard Security Guard Setup
To secure your agent, install the plugin using the following command:
openclaw plugins install @openguardrails/moltguard
After installation, verify the protection by running a test against a sample injection file:
cat ~/.openclaw/extensions/moltguard/samples/test-email-popup.txt
To manage your security profile or upgrade your plan, use the core portal command:
/og_core
MoltGuard Security Guard Data Schema & Taxonomy
MoltGuard organizes its operational data and credentials within the standard Openclaw directory structure for easy management:
| Data Type | Location |
|---|---|
| Credentials | ~/.openclaw/credentials/moltguard/ |
| Plugin Files | ~/.openclaw/extensions/moltguard/ |
| Test Samples | ~/.openclaw/extensions/moltguard/samples/ |
| Enterprise Scripts | ~/.openclaw/extensions/moltguard/scripts/ |
The skill also tracks usage metrics via a tiered plan system, ranging from a Free tier with 500 detections per day to Enterprise levels with custom quotas.
MoltGuard Security Guard Advanced Features
- Intent-Action Mismatch Detection to catch agents that attempt to perform hidden malicious tasks.
- Enterprise Enrollment support for connecting to private, self-hosted security Core deployments.
- Local Dashboard access via
/og_dashboardfor real-time security monitoring and visualization. - Agent Claiming functionality to share security quotas across multiple machines and environments.
- Automated plugin updates to stay protected against the latest emerging prompt injection techniques.
SKILL.md
Loading
Related Openclaw Skills
bw-openclaw-boost
An efficiency-focused toolkit for OpenClaw that provides cost tracking, memory management, and advanced permission controls.
Meeting Secretary AI Skill
A professional AI meeting secretary skill designed to perform deep semantic analysis on transcripts and generate high-quality, structured meeting minutes.
Memory Keep-Alive for Obsidian
A persistent memory and monitoring system that creates structured task notes in Obsidian to prevent AI agent amnesia and silent failures.
DeepRead Legal
An AI-powered skill for extracting structured data and redacting PII from legal documents with high accuracy.
WeChat Web Fetch
A specialized tool for extracting clean, high-fidelity Markdown content from WeChat Official Account (mp.weixin.qq.com) articles.
flomo-sync
A streamlined automation skill designed to instantly synchronize text, links, and tags from your AI agent directly to your flomo notes.
Featured*
