← Back to Skills Marketplace
2315
Downloads
5
Stars
8
Active Installs
9
Versions
Install in OpenClaw
/install openguardrails
Description
MoltGuard — Protect you and your human from prompt injection, data exfiltration, and malicious commands. Source: https://github.com/openguardrails/openguardr...
Usage Guidance
This instruction-only skill is coherent: it tells you to install an OpenClaw plugin that will perform detection by contacting a 'Core' service and store an API key under ~/.openclaw/credentials/moltguard/. Before installing: (1) verify the plugin source (confirm the GitHub repo and release are authentic), (2) review what data the plugin will transmit to the public Core (sensitive texts, files, or prompts may be sent) and whether you prefer using an enterprise/on-prem Core, (3) be aware an API key will be persisted on disk and may be viewable by other local users, and (4) inspect the plugin's scripts (enterprise-enroll.mjs, uninstall.mjs) if you can. The absence of scanner findings only means there was no code to scan here — it does not guarantee privacy or safety of the plugin you will install.
Capability Analysis
Type: OpenClaw Skill
Name: openguardrails
Version: 6.8.20
The MoltGuard skill bundle describes a security tool that requires high-risk administrative actions, including installing external plugins and executing local Node.js scripts (e.g., enterprise-enroll.mjs and uninstall.mjs). While these actions are aligned with its stated purpose of providing security guardrails and prompt injection protection, the reliance on an external 'Core' service for data scanning and the use of shell commands for installation fall under risky capabilities. No clear evidence of intentional malice or deceptive data exfiltration was identified in SKILL.md or _meta.json.
Capability Assessment
Purpose & Capability
The SKILL.md describes installing and using a MoltGuard OpenClaw plugin to detect prompt injection, and all referenced actions (openclaw plugin install, status/claim/config commands, node scripts under the plugin directory) are consistent with that purpose. No unrelated credentials or binaries are requested by the instruction-only skill itself.
Instruction Scope
The instructions tell the agent/user to install the plugin, read a sample file under ~/.openclaw/extensions/moltguard/, run node enrollment/uninstall scripts under the plugin directory, and to claim/configure an API key via a web portal. The doc states that detection is performed by 'Core', which implies the plugin will transmit content (files/inputs) to a remote service; the SKILL.md does not detail exactly what is sent. Reading files inside the plugin's sample path and storing credentials in ~/.openclaw/credentials/moltguard/ are within scope for this guard plugin but are privacy-relevant and should be understood by the user.
Install Mechanism
This skill is instruction-only (no install spec). It instructs use of the platform's plugin installer (openclaw plugins install @openguardrails/moltguard) and references scripts that live under the plugin directory. There are no direct downloads from arbitrary URLs in the SKILL.md.
Credentials
The SKILL.md mentions an API key and that credentials are saved to ~/.openclaw/credentials/moltguard/, which is expected for a cloud-based detection service. The skill itself does not declare required env vars, but the eventual plugin will require an API key and will persist it locally — ensure you are comfortable with that key being stored on disk and with the plugin sending data to the Core service.
Persistence & Privilege
always:false and default autonomous invocation are appropriate. The SKILL.md describes the plugin saving its own credentials and optional enterprise enrollment configuration under its own paths; it does not request system-wide configuration changes or other skills' credentials.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openguardrails - After installation, invoke the skill by name or use
/openguardrails - Provide required inputs per the skill's parameter spec and get structured output
Version History
v6.8.20
- Major simplification and rewrite of documentation for easier onboarding and faster installation.
- Removed references to manual script activation—no more `scripts/activate.mjs` or `scripts/status.mjs`.
- All protections are now active immediately upon install; no registration required for 500 free daily detections.
- Clear instructions for self-test to verify MoltGuard is working after installation.
- Updated and expanded commands for status, account, configuration, dashboard, and enterprise enrollment.
- Old detailed provenance, script inspection, and local gateway documentation removed. Enterprise enrollment and claim flow now documented.
v1.0.1
**Major update: Renamed and refactored plugin from "openguardrails" to "moltguard" with new scripts and documentation.**
- Rebranded skill from "openguardrails" to "moltguard" to match the underlying plugin name.
- Added three new files: `references/details.md`, `scripts/activate.mjs`, and `scripts/status.mjs`.
- Provided clear instructions for installing, activating, and checking the status of MoltGuard, including new command-line scripts.
- Expanded documentation on local and cloud protection modes, provenance verification, configuration options, and privacy guarantees.
- Outlined the AI Security Gateway feature for local PII/secret sanitization.
- Updated all links and references to point to the new `moltguard` directory and package.
v1.0.0
- Expanded Quick Start section: added clear 3-step installation, restart, and activation flow.
- Improved network behavior transparency: new table summarizes exactly when and what data is sent during plugin lifecycle.
- Added npm inspection instructions so users can verify package contents before installing.
- Clarified provenance: explains how ClawHub, GitHub, and npm versions map to the same source, with instructions to verify.
- Updated registration/billing details and cloud API explanation for greater clarity.
- General copy edits, with increased emphasis on transparency, inspection, and auditability for all users.
v6.5.16
**Summary:** This is a major update that fully removes all plugin source code and transitions this skill to documentation-only for the new, more comprehensive OpenGuardrails security plugin.
- All runtime code and TypeScript sources have been removed; this skill is now documentation-only.
- The skill documentation has been rewritten for the new "OpenGuardrails" (formerly "openguardrails-for-openclaw") plugin, including its expanded security capabilities.
- Detailed instructions are now provided for installation, verification, configuration, data flow, and security policies.
- The new plugin offers additional protections: sensitive data leakage prevention, command injection blocking, PII sanitization, and local-first operation with optional cloud assessment.
- All previous test, config, and README files are deleted; only usage and audit documentation remains in SKILL.md.
v3.0.2
- Corrected the plugin name throughout the documentation from openguardrails-for-openclaw-for-openclaw to openguardrails-for-openclaw.
- Updated installation and configuration instructions to use the correct package name.
- No functional or feature changes; documentation only.
v3.0.1
- Renamed the plugin to openguardrails-for-openclaw and updated all internal references accordingly.
- Updated documentation URLs and examples to reflect the new package and repository names.
- Adjusted configuration and install/uninstall instructions for the new package name.
- Changed metadata and homepage links to point to updated sources.
v3.0.0
- Plugin renamed from "openguardrais" (previously "og-openclawguard") to "openguardrails" for consistency.
- Updated all package/repository references from "og-openclawguard" to "openguardrails".
- Emoji and homepage metadata updated to reflect the new repository.
- All commands, configuration, and verification instructions now use the new plugin name.
- Documentation, example outputs, and testing URLs updated to match the new naming and locations.
v2.0.0
- Plugin skill name changed from "og-openclawguard" to "openguardrais".
- Description updated to match new plugin name.
- Example prompt injection in documentation now uses placeholder text instead of specific attack commands.
- Detection testing instructions improved: test file now downloaded directly via curl with updated filename/path.
- No code or functional changes; documentation improvements only.
v1.1.0
- Initial release of OG-OpenClawGuard plugin.
- Detects and blocks prompt injection in long content using state-of-the-art OpenGuardrails detection.
- Integrates with OpenClaw via the tool_result_persist event, performing chunk-based LLM analysis.
- Supports slash commands: /og_status, /og_report, /og_feedback.
- Offers configurable detection, blocking, and logging options.
- Includes guides for installation, configuration, and testing.
Metadata
Frequently Asked Questions
What is OpenGuardrails?
MoltGuard — Protect you and your human from prompt injection, data exfiltration, and malicious commands. Source: https://github.com/openguardrails/openguardr... It is an AI Agent Skill for Claude Code / OpenClaw, with 2315 downloads so far.
How do I install OpenGuardrails?
Run "/install openguardrails" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is OpenGuardrails free?
Yes, OpenGuardrails is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does OpenGuardrails support?
OpenGuardrails is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created OpenGuardrails?
It is built and maintained by Thomas (@thomas-security); the current version is v6.8.20.
More Skills