SkillGuard is an AI-driven security auditing tool designed to scan Openclaw Skills for malicious behavior and vulnerabilities before they are installed or executed.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install farnwick-skillguard
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 farnwick-skillguard 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 SkillGuard?
SkillGuard provides a critical layer of defense for developers and users of AI agents by analyzing skill source code for potential threats in real-time. As a specialized security utility for Openclaw Skills, it leverages advanced large language models to detect dangerous patterns such as credential theft, data exfiltration, and unauthorized persistence mechanisms. This ensures that every new capability you add to your workspace is safe, verified, and free from obfuscated malicious payloads.
By integrating directly into the skill management lifecycle, SkillGuard mitigates the risks inherent in executing third-party code. It evaluates the technical logic of shell scripts, configuration files, and metadata, providing a clear risk verdict ranging from clean to high alert. This proactive approach to security makes it an essential tool for anyone building or maintaining a production-ready ecosystem of Openclaw Skills.
SkillGuard Use Cases
- Vetting third-party Openclaw Skills before installing them on your local system or server.
- Periodically auditing your existing library of Openclaw Skills for security regressions or hidden vulnerabilities.
- Reviewing local skill development projects during the coding phase to ensure no accidental security leaks exist.
- Automating security gatekeeping within an AI agent's internal workflow when it attempts to fetch new capabilities.
How SkillGuard Works
- The tool downloads the target skill into a secure, temporary directory to isolate it from the main system.
- It recursively scans the directory, skipping binary files and truncating large files to optimize the AI analysis process.
- The source code and script logic are sent to a configured LLM provider like Anthropic or DeepSeek for a deep semantic security audit.
- SkillGuard evaluates the code against a checklist of threats, including unauthorized API key access, reverse shells, and package smuggling.
- A detailed risk report is generated, providing a definitive verdict (CLEAN, LOW, MEDIUM, or HIGH) and highlighting specific line-item warnings.
- For the install command, the tool pauses and prompts the user for final manual confirmation based on the audit results.
SkillGuard Setup
Ensure you have Python 3.6 or higher installed on your system. You will also need an API key for Anthropic, OpenRouter, or DeepSeek configured within your environment. Install the skill using the following commands:
# To scan and install a new skill
skillguard install <skill-name>
# To audit all currently installed Openclaw Skills
skillguard audit
# To scan a specific directory
skillguard scan /path/to/skill
SkillGuard Data Schema & Taxonomy
SkillGuard organizes its security findings into a structured report based on risk severity and threat category. It specifically monitors files within Openclaw Skills for access to sensitive system paths and network activity.
| Risk Level | Meaning |
|---|---|
| CLEAN | No security issues detected; safe to proceed. |
| LOW | Minor concerns identified, generally considered safe. |
| MEDIUM | Review recommended; contains suspicious patterns. |
| HIGH | Dangerous behavior detected; installation strongly discouraged. |
The scanning engine targets specific metadata and script types, including JSON configuration files, bash scripts, and Python modules, looking for unauthorized access to ~/.ssh/, ~/.openclaw/, and environmental variables.
SkillGuard Advanced Features
- Integration with high-performance LLMs like Claude Opus for sophisticated detection of logical vulnerabilities in Openclaw Skills.
- Automated audit mode that bulk-scans all installed skills across multiple system directories like /usr/lib/node_modules/ and user workspaces.
- Local development scanning mode allowing creators to verify their own Openclaw Skills before sharing them with the community.
- Intelligent truncation and binary skipping to handle large skill packages efficiently without losing security context.
SKILL.md
Loading
Related Openclaw Skills
Bookmark Organizer
Converts browser bookmark HTML exports into a clean, categorized, and searchable Markdown knowledge base with dead-link detection.
Exa
A neural search engine skill that provides AI agents with deep web context, code documentation, and research data.
Hello World Greeting Skill
A foundational greeting skill designed to demonstrate environment variable configuration and personalized automated responses.
English Tutor
A personalized AI-powered tutor designed to help users master American English through level-specific lessons and daily automated exercises.
Farseek
An AI-driven job search engine that matches developer skills to thousands of live job boards and ATS platforms.
S.H.I.T Journal Paper Recommender
A satirical AI agent skill that fetches and recommends humorous academic papers from the Sciences Humanities Information Technology Journal.
Featured*
