A mandatory security auditing and quarantine system designed to evaluate third-party Openclaw Skills for safety and compliance.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agxntsix-skill-auditor
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 agxntsix-skill-auditor 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 Auditor?
Skill Auditor acts as the primary security layer for any environment utilizing Openclaw Skills. Its core mission is to ensure that no third-party skill is integrated into a production system without first passing a rigorous, automated safety evaluation. By implementing a quarantine-first protocol, it isolates untrusted code in a temporary directory where it can be inspected for malicious patterns, unauthorized network calls, and potential security vulnerabilities.
This skill provides developers and system administrators with the tools needed to maintain a high trust environment. It doesn't just scan for viruses; it performs deep analysis on shell commands, environment variable access, and filesystem escape attempts. By generating clear risk scores ranging from CLEAN to CRITICAL, Skill Auditor empowers users to make informed decisions about the safety of new Openclaw Skills before they are granted access to sensitive local data.
Skill Auditor Use Cases
- Validating the safety of third-party Openclaw Skills sourced from ClawHub or external git repositories.
- Automatically triggering a security gate before any
clawhub installoperation is completed. - Reviewing existing skills for potential prompt injection vulnerabilities or hidden network requests.
- Implementing a standardized security audit workflow within a multi-agent AI system.
How Skill Auditor Works
- The auditor first clones or copies the target skill into a dedicated quarantine directory to prevent direct execution.
- A Python-based audit script scans the entire directory structure, including code files, configuration, and documentation.
- The system checks for suspicious indicators such as base64-encoded payloads, obfuscated imports, and dangerous shell commands.
- It evaluates the SKILL.md file against known prompt injection signatures and unauthorized permission requests.
- A final risk score is calculated based on findings, resulting in a specific exit code that allows or blocks the installation process.
Skill Auditor Setup
Skill Auditor should be located within your local skills directory. Ensure you have Python 3 available in your environment.
To perform a manual audit on a specific skill directory:
python3 skills/skill-auditor/scripts/audit_skill.py /path/to/skill-dir --human
To utilize the full quarantine and installation safety workflow:
bash skills/skill-auditor/scripts/quarantine.sh /path/to/skill-source
Skill Auditor Data Schema & Taxonomy
The skill utilizes a structured approach to identifying threats and reporting findings. It relies on internal reference catalogs to stay updated on emerging patterns in Openclaw Skills.
| Component | Description |
|---|---|
scripts/audit_skill.py |
The main engine responsible for scanning and risk scoring. |
references/known-patterns.md |
A database of known malicious code patterns and attack vectors. |
references/prompt-injection-patterns.md |
Specific signatures used to detect prompt injection attempts. |
| Audit Report | A JSON-formatted output containing detailed findings for every scanned file. |
Skill Auditor Advanced Features
- Quarantine-first architecture to ensure no untrusted code touches production directories until verified.
- Detailed risk rating system (CLEAN, LOW, MEDIUM, HIGH, CRITICAL) for granular security control.
- Comprehensive scanning for obfuscation techniques, including hex/unicode escapes and string manipulation tricks.
- Detection of suspicious network calls and filesystem escape patterns that could compromise the host machine.
- Support for both human-readable text summaries and machine-parseable JSON reports for CI/CD integration.
SKILL.md
Loading
Related Openclaw Skills
Research Logger
An automated AI research pipeline that logs search results from Perplexity into a structured SQLite database with full Langfuse observability.
Model Intel
A dynamic utility for querying real-time OpenRouter data to find, compare, and price Large Language Models on the fly.
Deep Search
Deep Search is a sophisticated search routing engine for Openclaw Skills that leverages Perplexity AI models to provide tiered web research capabilities.
Agent Memory
A sophisticated intelligence layer providing vector search, knowledge graphs, and structured data storage for AI agents.
AI CFO
A full AI Chief Financial Officer that integrates banking and payment data into real-time business intelligence and automated reporting.
Alchemy
A high-performance interface for AI agents to query blockchain data, NFT metadata, and token balances using the Alchemy API.
Featured*
