A comprehensive security auditing tool designed to scan, detect, and neutralize threats within Openclaw Skills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skillguard-scanner
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 skillguard-scanner 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 serves as a critical security layer for developers and users of AI agents. In an era where third-party integrations can introduce significant vulnerabilities, this tool provides automated static analysis to identify malicious patterns before they can compromise your system. By scanning Openclaw Skills for everything from hardcoded credentials to sophisticated reverse shells, SkillGuard ensures that your automated workflows remain secure and private.
The tool is particularly vital given the lack of official vetting in many agent ecosystems. Whether you are auditing your own code or verifying external contributions, SkillGuard provides a transparent risk assessment. It analyzes the technical structure of Openclaw Skills to find hidden threats like code obfuscation, data exfiltration scripts, and memory poisoning instructions that could lead to unauthorized system access.
SkillGuard Use Cases
- Scanning new Openclaw Skills from remote repositories like ClawHub prior to installation to prevent supply chain attacks.
- Performing routine security audits of all installed Openclaw Skills to detect any unauthorized changes or newly discovered vulnerabilities.
- Identifying typosquatting attempts where malicious skills use names similar to popular Openclaw Skills to trick users into installation.
- Checking for credential theft patterns where a skill might attempt to access .env files, API tokens, or SSH keys without authorization.
How SkillGuard Works
- The scanner is initiated via the command line, targeting either a local directory or a remote skill identifier.
- SkillGuard performs a deep inspection of the source code, searching for critical threat signatures like bash reverse shells or obfuscated payloads.
- The tool audits the skill's metadata and documentation for suspicious prerequisites or instructions that could poison the agent's long-term memory.
- Each detected issue is assigned a risk score and a false-positive likelihood estimate based on the context of the code.
- A comprehensive report is generated, categorizing the safety of the Openclaw Skills into five distinct risk levels from Clean to Critical.
SkillGuard Setup
To begin securing your environment, ensure you have the necessary scripts in your path. You can run the following commands to audit your Openclaw Skills:
# Audit all installed skills at once
python3 {scripts}/scanner.py
# Scan a specific skill before installation from ClawHub
python3 {scripts}/scanner.py --fetch-clawhub <skill-name>
# Check for typosquatting on a specific name
python3 {scripts}/scanner.py --check-name <name>
SkillGuard Data Schema & Taxonomy
SkillGuard organizes its security findings into a structured format for easy review. The primary output is a JSON report that details the security posture of your Openclaw Skills.
| Data Point | Description |
|---|---|
| Risk Level | Visual indicator (🔴 to ✅) based on the calculated security score. |
| Threat Category | Classification of the risk (e.g., Code Obfuscation, Credential Access). |
| FP Likelihood | An estimate (low/medium/high) of whether the finding is a false positive. |
| scan_results.json | The default export file containing detailed metadata for every audited skill. |
SkillGuard Advanced Features
- Multi-vector threat detection covering malware, crypto-drainers, and prompt injection specifically for Openclaw Skills.
- Intelligent typosquatting detection using Levenshtein distance to catch deceptive skill naming conventions.
- Remote fetch capabilities allowing for "sandbox-style" auditing of skills before they are added to the local environment.
- Automated detection of memory poisoning attempts targeting core agent files like SOUL.md and AGENTS.md.
SKILL.md
Loading
Related Openclaw Skills
LegalFrance
An advanced AI legal assistant providing hybrid retrieval and verifiable citations for French consolidated laws within the Openclaw Skills ecosystem.
ChromaDB Memory
A 100% local, self-hosted long-term memory system using ChromaDB and Ollama for intelligent context injection.
OpenClaw Default Agent Backstory
A systematic framework for bootstrapping and maintaining stable AI agent identities through guided interviews and core context file generation.
Uptime Kuma Monitoring
A powerful CLI-driven skill to manage Uptime Kuma monitoring instances, allowing for automated status checks and monitor configuration.
mcporter
A powerful CLI utility for interacting with Model Context Protocol (MCP) servers, enabling direct tool calling, authentication, and code generation.
OpenClaw Ticket Assistant
An intelligent assistant that automates the creation of support tickets by analyzing customer service chat screenshots.
Featured*
