skill-auditor

What This Skill Does

The Skill Auditor is a powerful security analysis tool designed to protect your OpenClaw agent from malicious code hidden within third-party skills. By acting as a firewall between your environment and the skills you install, it scans for common threat vectors, including credential theft, prompt injection attempts, and obfuscated backdoors. While the core functionality is ready out-of-the-box, advanced users can leverage optional AST (Abstract Syntax Tree) dataflow analysis to track how data moves through your agent's execution path. It is an essential utility for users who prioritize privacy and system integrity when extending their AI agent's capabilities.

Installation

To install the auditor, run the following command in your terminal: clawhub install openclaw/skills/skills/rubenaquispe/skill-auditor. Once installed, it is highly recommended to run the interactive setup wizard located at cd skills/skill-auditor and executing node scripts/setup.js. This wizard will automatically detect your operating system, verify Python requirements for AST analysis, and configure auto-scan preferences. If you prefer a non-interactive approach, you can manually enable features using the CLI flags provided in the documentation.

Use Cases

• Vetting new skills: Automatically audit every new skill you download from the hub before allowing execution.
• Scheduled Audits: Periodically run node scripts/audit-installed.js to ensure that existing skills have not been compromised through updates.
• Deep Dataflow Analysis: Use the AST analysis feature to investigate how a skill handles sensitive information like your API keys or personal files, ensuring no data is exfiltrated to external endpoints.
• Security Compliance: Maintain a secure workspace by filtering out skills that do not meet your defined security thresholds or risk levels.

Example Prompts

1. "Auditor, scan the latest skill I just installed in the downloads directory for potential prompt injection risks."
2. "Run a full system audit of all currently installed skills and save the results to a JSON file for my review."
3. "Show me a report of all skills that have critical severity risks or require sensitive system permissions."

Tips & Limitations

To get the most out of this tool, ensure Python 3.8+ is installed on your system if you intend to use the advanced AST dataflow analysis. While the core scanner works on all platforms with standard Node.js, the AST analysis is significantly more effective at detecting hidden, obfuscated logic. Remember that no security scanner is foolproof; always exercise caution when granting permissions to third-party code. If a skill is flagged as 'CRITICAL', do not execute it until you have manually inspected the source code.