Description

Security researchers have uncovered a malicious extension on the Visual Studio Code Marketplace that impersonated an AI-powered coding assistant associated with the Moltbot project. Distributed under the name “ClawdBot Agent-AI Coding Assistant,” the extension was published in late January 2026 and was designed to covertly deploy malware after installation. Once executed, it established persistent remote access on affected machines using a modified instance of ConnectWise ScreenConnect, enabling attackers to control compromised developer environments. Analysis by researchers at Aikido Security revealed that the extension represented a supply-chain style threat, packaged to appear legitimate and trustworthy. Although Moltbot itself does not provide an official VS Code plugin, the attackers created a convincing listing complete with polished visuals and claims of compatibility with well-known AI platforms such as OpenAI, Anthropic, and Google. This presentation increased the likelihood that developers would install the extension without verifying its authenticity. After installation, the extension executed automatically whenever VS Code launched. It retrieved configuration data from an attacker-controlled domain, which directed the download of payload components including an executable masquerading as Code.exe. This binary functioned as a preconfigured ScreenConnect client that connected to a remote relay server, silently opening a persistent control channel. To maintain reliability, the malware incorporated fallback delivery methods. These included sideloading a malicious DWrite.dll loader written in Rust or using scripted downloads from alternate infrastructure if the primary server became unavailable. The campaign highlights the growing risk posed by malicious packages in trusted developer ecosystems and reinforces the importance of validating publisher authenticity before installing extensions.