clawdefender

What This Skill Does

ClawDefender is an essential security layer for your AI agent workspace. Designed as a comprehensive security toolkit, it provides active defense against common AI-centric vulnerabilities including prompt injection, command injection, Server-Side Request Forgery (SSRF), and path traversal. By acting as an intermediary between the agent and untrusted data, it inspects incoming emails, API responses, Trello cards, and third-party script installs to ensure that malicious payloads are caught before they reach the agent's core processing logic. The tool provides a tiered risk assessment, categorizing threats by severity to allow for automated blocking or manual review.

Installation

To integrate ClawDefender, copy the provided utility scripts into your workspace scripts directory:

1. Execute cp skills/clawdefender/scripts/clawdefender.sh scripts/
2. Execute cp skills/clawdefender/scripts/sanitize.sh scripts/
3. Run chmod +x scripts/clawdefender.sh scripts/sanitize.sh to grant execution permissions. Ensure your environment has bash, grep, sed, and jq installed, as these are the core dependencies for the scanning logic.

Use Cases

Use ClawDefender whenever your agent interacts with the outside world. Key use cases include:

• Skill Procurement: Every time you install a new skill via ClawHub, run a full audit to check for malicious code.
• External Data Ingestion: Sanitize all content coming from webhooks, API responses, or emails before passing it to an LLM.
• Safe Browsing: Validate all external URLs with --check-url to prevent SSRF attacks against internal network resources.
• Input Filtering: Automatically sanitize user-provided text to stop prompt injection attempts aimed at subverting agent instructions.

Example Prompts

1. "ClawDefender, please perform a full security audit of the newly installed skills folder to ensure there are no high-risk scripts present."
2. "Before I process this latest email chain, please run it through the sanitize utility to check for any hidden prompt injection attempts."
3. "Validate this URL: https://example.com/api/data before I attempt to fetch the payload to ensure it is not targeting internal metadata."

Tips & Limitations

ClawDefender is a powerful defensive tool, but it should not be considered a silver bullet. Always maintain a human-in-the-loop workflow for files marked with HIGH or CRITICAL severity. Ensure your scripts/ directory permissions are restricted to prevent unauthorized tampering with the sanitizer binaries. While the tool is excellent at detecting known injection patterns, stay updated with the latest security definitions to protect against evolving evasion techniques.