OpenClaw Partners with VirusTotal to Enhance Security Against Malicious Skills

OpenClaw, previously known as Moltbot and Clawdbot, has announced a collaboration with VirusTotal to scan skills uploaded to its ClawHub marketplace. This partnership aims to strengthen security within the agentic ecosystem, especially after reports highlighted the presence of hundreds of malicious skills. The scanning process involves creating a unique SHA-256 hash for each skill and checking it against VirusTotal’s database, utilizing their new Code Insight capability for further analysis.

Skills receiving a ‘benign’ verdict from Code Insight are automatically approved, while suspicious ones are flagged, and malicious skills are blocked from download. OpenClaw also conducts daily re-scans of active skills to identify any that may have turned malicious over time. However, OpenClaw’s maintainers caution that this scanning is not foolproof, as some malicious skills may evade detection.

The urgency for this partnership is underscored by findings from various analyses revealing that many skills masquerade as legitimate tools but contain hidden malicious functionalities. These can include data exfiltration, backdoor injections, and the installation of malware. The potential for AI agents to act as covert data-leak channels poses significant risks, especially as these agents gain access to sensitive user data.

Security concerns have escalated with OpenClaw’s rising popularity, particularly as it is deployed on employee endpoints without formal IT approval. This creates a new class of Shadow AI risk, where unauthorized tools can facilitate data movement and network connectivity outside standard security controls. The Chinese Ministry of Industry and Information Technology has even issued alerts regarding misconfigured instances of OpenClaw, urging users to implement protective measures.

Security Issues Identified

Several glaring security issues have been identified in OpenClaw, including the storage of credentials in cleartext and the use of insecure coding patterns. A recent analysis of 3,984 skills on ClawHub found that 283 skills, approximately 7.1%, contained critical security flaws that could expose sensitive credentials. Additionally, a zero-click attack has been documented, allowing attackers to plant backdoors through seemingly harmless documents processed by the AI agent.

As OpenClaw continues to evolve, it is crucial for users and organizations to remain vigilant. The integration of VirusTotal scanning is a step forward, but it is essential to recognize that the risk landscape is continually changing. Users should be proactive in monitoring their environments and ensuring that they are not inadvertently exposing sensitive data through these AI agents.