OpenClaw integrates VirusTotal scanning to detect malicious ClawHub skills

OpenClaw (formerly Moltbot and Clawdbot) announced that it is partnering with VirusTotal, owned by Google, to scan skills uploaded to ClawHub, its skills marketplace, as part of broader efforts to strengthen the security of the agent ecosystem.

See also: Moltbook: The new social network for AI that threatens to wipe out humanity

“All skills published on ClawHub are now scanned using VirusTotal intelligence, including the new Code Insight feature,” said OpenClaw founders Peter Steinberger, Jamieson O’Reilly , and Bernardo Quintero. “This provides an additional layer of security for the OpenClaw community.”

The process involves generating a unique SHA-256 hash for each skill and cross-checking it against the VirusTotal database for a match. If not found, the skill pack is uploaded to the malware scanner for further analysis using VirusTotal Code Insight.

Skills that have a “benign” Code Insight verdict are automatically approved by ClawHub, while those that are deemed suspicious are flagged with a warning. Any skill deemed malicious is excluded from download. OpenClaw also said that all active skills are rescanned daily to detect scenarios where a previously clean skill becomes malicious.

However, OpenClaw maintainers warned that VirusTotal scanning is not a "panacea" and that there is a possibility that some malicious skills that use cleverly hidden command injection payloads could go unnoticed.

See also: OpenClaw: Researchers discover 341 malicious ClawHub skills

In addition to partnering with VirusTotal, the platform is expected to publish a comprehensive threat model, public security roadmap, formal security reporting process, and details on security auditing of its entire code.

This development follows reports that found hundreds of malicious skills on ClawHub, prompting OpenClaw to add a reporting option that allows logged-in users to flag a suspicious skill. Multiple analyses revealed that these skills pretend to be legitimate tools but hide malicious functionality to extract data, inject backdoors for remote access, or install theft malware.

The recent viral popularity of OpenClaw and Moltbook, a social network where autonomous AI agents built on OpenClaw interact with each other on a Reddit-like platform, has raised security concerns.

While OpenClaw acts as an automation engine to enable workflows, interact with web services, and operate across devices, the access given to skills, combined with the fact that they can process data from untrusted sources, can open the door to risks such as malware and command injection.

See also: OpenClaw flaw allows remote code execution

In other words, the integrations, while convenient, significantly broaden the attack surface and expand the set of untrusted inputs consumed by the agent, turning it into a “Trojan horse” for data extraction and other malicious actions. Backslash Security has described OpenClaw as “AI With Hands.”