Hackers rely on OpenClaw's AI Agent "Skill."..

Hackers rely on OpenClaw's AI Agent "Skill" sending malware down the victim's system.

Artificial intelligence assistants, or AI Agents, can be called the breakthrough innovation of this decade to be a new step in its work, but today many AI Agents have been exploited in cyberterrorism to become a security controversy, and this time AI Agents have been in the news again.

According to a report by the website Cyber Security News, VirusTotal, a popular suspected file and URL authentication site, has detected a group of hackers using OpenClaw (formerly Clawdbot and Moltbot), an AI Agent for automating system management tasks. The AI Agent can handle both Shell Commands, File Operations and Network Requests to release malware through the Skills add-on, which is an add-on for various tasks. Automatically through Instruction, written on a file skill.md within the folder of each Skill add-on.

In order to use Skill, Skill works in the 3rd Party Code with complete system access to manage the system of the user using this AI Agent. Therefore, those who want to use it have to do a lot of steps to use Skill on the system. The user must paste and run the code through Terminal (or Run), download the Binary file and run the script as specified. To do this, the user must give a lot of trust to what is on the platform by hackers. It relies on this point to release malware into the victim's machine.

VirusTotal checked up to 3,016 Skill add-ons on the OpenClaw platform and found that more than a hundred Skills were malware-based when Google's AI malware analyzer, Gemini 3 Flash, was used to detect Skills add-ons by checking if they have dangerous capabilities, such as external code Execution, Sensitive Data access, and unsafe network management? Replace the use of analysis to verify specific features of malware (Virus Signature Analysis).

This allows us to divide these hazards into two groups:

Groups with poor security management, such as unsafe API implementation, hard-coded Secrets, and unsafe command execution.

Groups are created to deliberately attack victims with capabilities such as File Exfiltration, Remote Control, and Malware Installation.

And when it looked deeper, it was found that among all the developers of the Danger add-on, there were the developers who used the name "hightower6eu" that were the most prominent from the development and distributed up to 314 Danger add-ons that had many different capabilities, including Crypto Analytics, Finance Tracking, and Social Media Analysis. All of the add-ons were specified to download and run code from other sources outside (untrustworthy). Come to install first

For example, the "Yahoo Finance" add-on, when using the Anti-Virus detection tool, will find that the add-on is secure, but the add-on instructs the user to download a compressed file in Password Protected, which inside contains an openclaw-agent.exe file. This file, when verified, will be found to be a Trojan malware file.

While users of the macOS operating system are told to download a Shell Script file that is encrypted to bounce the system (Obfuscation) in Base64 from the glot [.] io website, which leads to a Binary file in the Mach-O file format. Inside the file is a malware type that steals data from the victim, or an Infostealer called AMOS (Atomic Stealer).

# Trending # Lemon 8 Howtoo # lemon 8 diary # openclaw # freedomhack