agentguard

What This Skill Does

GoPlus AgentGuard is a specialized security framework for OpenClaw AI agents, functioning as an automated security auditor. It acts as an active defense layer that monitors, analyzes, and guards against malicious code, vulnerabilities, and unauthorized access. By integrating AgentGuard, your agent gains the ability to introspect codebase structures, evaluate runtime action safety, manage trust levels for external skills, and maintain a verifiable audit log of security events. It is designed to proactively identify risks before they manifest into breaches.

Installation

To integrate this security guard into your environment, run the following command in your terminal: clawhub install openclaw/skills/skills/0xbeekeeper/security

Use Cases

AgentGuard is essential for developers and agent operators handling high-stakes workflows. Key use cases include:

• Third-Party Code Audit: Automatically scanning new skill repositories for obfuscated code, remote loaders, or shell execution risks before execution.
• Runtime Protection: Evaluating whether a requested tool action (like a file read or API call) deviates from the agent's expected safety profile.
• Secret Management: Preventing the accidental leakage of environment variables, SSH keys, or hardcoded private keys.
• Security Compliance: Auditing smart contracts (Solidity) for reentrancy, dangerous selfdestruct patterns, and unauthorized token transfers.

Example Prompts

1. "AgentGuard, scan the project directory at ./plugins and report any critical vulnerabilities found in the source files."
2. "I am about to execute this external library, check if the code contains any remote loading or shell execution patterns."
3. "Show me the recent audit log and change the current configuration to strict mode to block all unverified actions."

Tips & Limitations

• Depth of Analysis: AgentGuard performs deep static analysis, but it is not a substitute for human penetration testing. Always review critical security alerts manually.
• Performance: Scanning very large codebases may take significant time; use specific paths to narrow the focus.
• Pattern Matching: The tool relies on predefined security rules. Ensure you keep the agent updated to benefit from the latest detection signatures.
• Context Awareness: Markdown scanning is limited to code blocks to avoid false positives from documentation text, so ensure sensitive scripts are properly encapsulated.