skulk-skill-scanner

What This Skill Does

The skulk-skill-scanner is a specialized security tool designed for the OpenClaw ecosystem. It acts as a static analysis engine that scrutinizes OpenClaw skill directories to identify potentially malicious patterns before they are executed. By parsing source code, manifest files, and local dependencies, the scanner looks for red flags such as unauthorized data exfiltration, attempts to access credentials, malicious prompt injection vectors, and destructive shell commands. It serves as an essential gatekeeper for any user looking to integrate third-party functionality into their agent.

Installation

You can install this tool directly using the OpenClaw Hub CLI. Run the following command in your terminal: clawhub install openclaw/skills/skills/adainthelab/skulk-skill-scanner

Use Cases

• Safety Auditing: Run this against a new skill downloaded from the community to verify it does not contain hidden network calls or obfuscated code.
• Pre-Publishing Verification: If you are a developer, use this to scan your own projects to ensure you haven't accidentally left sensitive code patterns before sharing your work on ClawHub.
• CI/CD Integration: Integrate the scanner into your automated build pipelines using the --json or --summary flags. Since the tool returns an exit code of 1 on failure, it is perfect for stopping insecure code from ever reaching a deployment stage.

Example Prompts

• "Scan the folder ./skills/community-analytics-tool and let me know if it attempts to access my environment variables."
• "Run a silent check on my ./skills/my-new-plugin and output the results in JSON format for my report."
• "Audit the local skill at ./skills/web-scraper using the verbose flag so I can review every rule triggered by the static analyzer."

Tips & Limitations

Always remember that this scanner relies on pattern matching. While it is excellent at detecting known exploit signatures, it is not a silver bullet against advanced, runtime-based threats or highly sophisticated social engineering. Treat this as your primary layer of defense, but keep a cautious eye on any skill requesting high-level system permissions. If a skill achieves a 'Warn' or 'Fail' score, refrain from installing it until you have manually inspected the codebase. Finally, remember to regularly update the SAFE_DOMAINS array within the script to ensure your legitimate network requests aren't incorrectly flagged.