Openclaw Defender

What This Skill Does

Openclaw Defender serves as the primary security layer for the OpenClaw agent ecosystem, specifically designed to mitigate the risks associated with the skill supply chain attacks identified by Snyk in 2026. This comprehensive framework acts as a runtime shield, protecting your agent from prompt injection, credential exfiltration, and malicious memory poisoning. By implementing robust File Integrity Monitoring, the skill ensures that critical configuration files like SOUL.md remain untampered. Its advanced Runtime Protection layer intercepts suspicious network requests, validates command execution, and prevents unauthorized access to sensitive file paths. The system utilizes a zero-trust architecture, requiring manual approval for new skill installations and enforcing strict blocklists to neutralize known threats immediately. Furthermore, the integration of a Kill Switch provides an automatic fail-safe mechanism, halting operations instantly upon the detection of a compromise to ensure data integrity and system availability.

Installation

To begin, ensure you have the OpenClaw workspace environment active. The skill is installed via the package manager using the command: clawhub install openclaw/skills/skills/nightfullstar/openclaw-defender. Once installed, you must establish a security baseline to allow the system to recognize legitimate file hashes. Execute ./skills/openclaw-defender/scripts/generate-baseline.sh from your ~/.openclaw/workspace directory. Verify the generated hashes in .integrity/ to confirm they reflect your current, trusted setup. Finally, automate the security checks by adding the provided cron job to your system schedule, ensuring the agent remains monitored every ten minutes.

Use Cases

• Proactive Threat Mitigation: Protects agents in production environments from malicious scripts installed through third-party skill repositories.
• Forensic Auditing: Use the generated JSON Lines logs to analyze potential attack vectors and refine security policies.
• Data Exfiltration Prevention: Automatically blocks unauthorized network calls that attempt to move sensitive agent memory blobs to external endpoints.
• Secure Lifecycle Management: Ensures that all agents follow a mandatory human-in-the-loop review process for new automation features.

Example Prompts

1. "Openclaw Defender, please audit my current skill library and provide a report on any unauthorized file modifications detected in the last 24 hours."
2. "Run a full security scan on the system and confirm that the latest memory update hasn't triggered a false positive for code injection."
3. "Disable all non-essential outbound network requests until I review the current runtime logs for potential credential leaks."

Tips & Limitations

Always review the ~/.openclaw/logs/integrity.log file weekly to stay informed of system state changes. Be aware that the Collusion Detection feature requires your agent's execution path to wrap skill calls with runtime-monitor.sh start and end commands; without this manual integration, cross-skill monitoring will not function. Ensure that your whitelist is updated frequently if you utilize custom or private skills, as the zero-trust policy will block unlisted binaries by default.