An advanced security auditing tool designed to detect malicious behaviors and obfuscated code within AI agent skills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skill-security-reviewer
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 skill-security-reviewer 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 Skill Security Reviewer?
Skill Security Reviewer v3.0.0 is a specialized developer tool for auditing the safety and integrity of AI coding agent skills. Unlike traditional scanners, it adopts a user-centric security perspective, focusing on how a skill might interact with or attack the host environment once installed. By integrating this into your workflow for Openclaw Skills, you can identify hidden risks before they compromise your local system.
The tool features robust detection layers for encoding evasion, encryption, and dynamic code generation to expose hidden payloads. It provides deep visibility into what an agent skill will actually do once installed, protecting sensitive user data such as SSH keys, cloud credentials, and browser cookies. It is an essential component for any developer maintaining a secure environment for their Openclaw Skills.
Skill Security Reviewer Use Cases
- Auditing a newly downloaded skill before local installation to prevent supply chain attacks.
- Performing regular security health checks on existing Openclaw Skills in your directory.
- Identifying obfuscated or encrypted malicious payloads in suspicious or unknown extensions.
- Threat modeling and risk assessment for enterprise-grade AI agent deployments.
How Skill Security Reviewer Works
- The tool scans the local environment to locate the target skill directory and its associated SKILL.md and script files.
- It extracts code blocks, shell scripts, and configurations for deep content inspection.
- It performs Shannon entropy analysis to identify likely encrypted or compressed content blocks.
- Multi-layer de-obfuscation is applied to reveal hidden code within Base64, Hex, or Unicode escapes.
- The system executes 94 distinct security checks across categories like data theft, command execution, and prompt injection.
- A final risk score is calculated, incorporating obfuscation bonuses for intentionally hidden logic.
- A comprehensive markdown security audit report is generated with actionable recommendations.
Skill Security Reviewer Setup
To utilize the Skill Security Reviewer, ensure it is available in your agent's skill path. You can initiate a security audit on any of your Openclaw Skills by using the following command structure:
/skill-security-reviewer {target-skill-name}
Example for auditing a suspicious skill:
/skill-security-reviewer suspicious-obfuscated-skill
Skill Security Reviewer Data Schema & Taxonomy
The skill organizes its analysis into a structured directory and generates detailed markdown reports. The data is categorized based on the following taxonomy:
| Category | Description | Detection Items |
|---|---|---|
| OBFUSCATION | Evasion and hiding techniques | 41 items (Encoding, Encryption, Entropy) |
| THREATS | Explicit malicious behaviors | 53 items (Theft, Exec, Persistence, Exfil) |
| SCORING | Risk determination metrics | 0-100 scale with severity weightings |
Reports are saved at ./{target-skill-name}-review-report/report-{YYYYMMDD-HHMMSS}.md.
Skill Security Reviewer Advanced Features
- Entropy Analysis: Automatically flags high-entropy strings that likely contain encrypted payloads or compressed binaries.
- Multi-layer De-obfuscation: Recursively decodes nested layers of encoding, such as Base64 strings nested inside Hex-encoded variables.
- Anti-Analysis Detection: Identifies code specifically designed to detect debuggers, virtual machines, or sandbox environments.
- Dynamic Code Reconstruction: Simulates string concatenation and splitting to uncover hidden commands like curl or rm -rf.
- Enhanced Scoring Engine: Applies weighted bonuses for skills within the Openclaw Skills ecosystem that use evasion techniques to hide their intent.
SKILL.md
Loading
Related Openclaw Skills
Clown Suit (clown-suit)
An operational framework for AI agents that prioritizes direct function and authentic pattern recognition over corporate image management.
openclaw-direct
A lightweight, secure, one-click installer for the OpenClaw ecosystem that bypasses the need for Docker.
CNINFO Announcement Scraper
An automated scraper and analyzer for extracting positive market catalysts from official CNINFO A-share disclosures.
CLS News Scraper
An automated engine for extracting actionable stock market catalysts and real-time news from Cailian Press (CLS).
Immich API Bridge
A powerful REST API bridge for managing self-hosted Immich photo instances, enabling seamless automation for media assets and libraries.
Writing Inspiration
A heuristic writing assistant that helps users brainstorm, structure, and draft articles through interactive, step-by-step questioning.
Featured*
