A comprehensive runtime security guard for OpenClaw agents providing local-first protection against exfiltration, injection, and sensitive data leaks.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install openguardrails-security
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 openguardrails-security 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?
MoltGuard is an open-source, fully auditable runtime security plugin developed by OpenGuardrails specifically for OpenClaw agents. It serves as a critical defense layer that monitors agent behavior to prevent data exfiltration, credential theft, and command injection. By adding this plugin to your collection of Openclaw Skills, you can ensure that your automated agents operate within safe boundaries, protecting both your local file system and your sensitive credentials.
The plugin is designed with a local-first philosophy, meaning high-assurance protections like blocking shell escapes and redacting prompt injections happen entirely on your machine. For users seeking deeper behavioral analysis, MoltGuard offers an optional cloud-based assessment that identifies complex intent-action mismatches without ever sharing the actual content of your messages.
MoltGuard Use Cases
- Prevent unauthorized reading of sensitive local files followed by suspicious network calls.
- Block malicious shell command injections (e.g., via backticks or pipe symbols) within tool parameters.
- Sanitize PII, credit card numbers, and API keys from outgoing LLM requests using the local security gateway.
- Monitor for unusual tool sequences that might indicate an agent has been compromised by prompt injection.
How MoltGuard Works
- The plugin intercepts tool calls made by the agent to perform immediate local risk assessment based on predefined security patterns.
- If a high-risk action is detected, such as a shell escape or unauthorized data access, the action is blocked before execution.
- For activated users, behavioral metadata (tool names and sequence) is sent to the OpenGuardrails core for advanced threat detection.
- The integrated AI Security Gateway acts as a local proxy, scrubbing sensitive data from the request stream before it reaches the model provider.
MoltGuard Setup
To install and activate MoltGuard, execute the following commands in your terminal:
# 1. Install the plugin package
openclaw plugins install @openguardrails/moltguard
# 2. Restart the gateway to load the security plugin
openclaw gateway restart
# 3. Register the agent and generate a claim URL
node {baseDir}/scripts/activate.mjs
After running the activation script, visit the provided claim URL to complete your registration and receive your API key.
MoltGuard Data Schema & Taxonomy
MoltGuard manages security data and configurations across the following structure:
| Component | Path / Format | Description |
|---|---|---|
| Credentials | ~/.openclaw/credentials/moltguard/credentials.json |
Stores the unique API key used for cloud behavioral analysis. |
| Configuration | ~/.openclaw/openclaw.json |
Contains plugin settings like blockOnRisk and agentName. |
| Security Gateway | http://127.0.0.1:8900 |
A stateless local proxy for real-time PII and secret sanitization. |
| Source Code | scripts/ |
Contains activate.mjs and status.mjs for lifecycle management. |
MoltGuard Advanced Features
- Local-only mode provides full protection against exfiltration and injection without any outbound network calls.
- Real-time PII sanitization via the AI Security Gateway supports SSNs, IBANs, phone numbers, and more.
- Configurable
blockOnRisksetting allows users to toggle between strict enforcement and observation-only modes. - Support for custom API endpoints and dashboard URLs for enterprise-grade monitoring and reporting.
- Audit-friendly architecture allows users to verify the npm package against the open-source GitHub repository before installation.
SKILL.md
Loading
Related Openclaw Skills
flaw0
An AI-powered security scanner designed to detect vulnerabilities in OpenClaw code, plugins, and Node.js dependencies.
OpenGuardrails
OpenGuardrails is an open-source runtime security plugin designed to protect AI agents from data exfiltration, prompt injection, and sensitive data leakage.
Task Dispatcher
An intelligent orchestration hub that analyzes, decomposes, and distributes complex tasks to specialized sub-agents while ensuring reliability and progress monitoring.
Flutter Testing & QA Automation
A comprehensive testing suite for Flutter applications covering unit, widget, and end-to-end integration tests.
Test Skill
A diagnostic utility for verifying the functionality and configuration of your AI agent environment.
ZTP (Zero Trust Protocol) Security Skill
A mandatory security gatekeeper that audits and validates new code, scripts, and MCP servers against the SEP-2026 Zero Trust protocol.
METADATA
Featured*
