Researchers Find 341 Malicious ClawHub Skills Stealing Data from OpenClaw Users
Researchers in the field of cybersecurity at Koi Security have uncovered a significant supply chain danger that is targeting users of OpenClaw, a well-known artificial intelligence assistant that you operate on your own machine. When it comes to the field of artificial intelligence research, this is an unexpected discovery. During the course of the audit, it was determined that an astounding 341 malicious ClawHub Skills had been uploaded to the official repository of the site. These abilities were developed with the intention of stealing confidential information such as API tokens, bitcoin keys, and login information.
As the use of artificial intelligence assistants such as OpenClaw (formerly known as Clawdbot and Moltbot) becomes increasingly mainstream, the discovery demonstrates that “agentic” security vulnerabilities are becoming more widespread. Individuals and businesses alike face a significant challenge in the form of malicious ClawHub Skills due to the fact that these agents are designed to operate with a great deal of access to the system.
Discovering 341 Unsound Skills on ClawHub
When researchers saw a curious spike in the amount of third-party “skills” or add-ons on ClawHub, they started looking into the matter. An OpenClaw bot with the name “Alex” was used by the researchers to conduct a thorough examination of all 2,857 skills that were available on the market. This bot was designed to seek for potential dangers. Approximately twelve percent of the repository was comprised of malicious ClawHub skills, which was a frightening discovery.
A total of 341 malicious ClawHub Skills were discovered, and 335 of those skills were linked to a single, meticulously organized campaign that was given the name ClawHavoc when it was discovered. In the majority of cases, this operation targeted machines running macOS and Windows. In order to spread sophisticated malware that steals information, it used people’s trust in open-source artificial intelligence ecosystems.
Researchers Find 341 Malicious ClawHub Skills Stealing Data from OpenClaw Users
Beware of Fake Dropbox Phishing Attack that Harvest Login Credentials
What type of phishing attack is the fake Dropbox attack?
A planned social engineering effort, known as the Fake Dropbox Phishing effort, is an attempt to dupe individuals into divulging their private login credentials in order to get access to Dropbox. This Fake Dropbox Phishing Attack does not use software vulnerabilities as its objective, as is the case with traditional hacking. In its place, it targets what is known as the “human element.” Attackers are able to fool individuals into visiting to fake websites that seem exactly like the real Dropbox login page by making them feel as though they have to act quickly and by replicating real communications.
After the year 2026, researchers have observed a significant increase in the number of these activities. The term “Business Email Compromise (BEC) 3.0” or “Living-off-Trusted-Sites” (LOTS) attacks is often used to refer to these types of activities. The use of these terms demonstrates how the Fake Dropbox Phishing Attack actually hosts malicious files by utilizing legitimate cloud providers such as Vercel,
Google, and even Dropbox itself. Because of this, it is extremely difficult for any of the standard email security filters to locate them.
Beware of Fake Dropbox Phishing Attack that Harvest Login Credentials
