Description

Cybersecurity researchers have identified an ongoing malware campaign named FakeGit that has created nearly 7,600 malicious GitHub repositories using approximately 6,600 fake or compromised-looking developer profiles. More than 800 repositories reportedly impersonated AI skills and Model Context Protocol (MCP) servers for popular tools and services. These repositories used copied projects, convincing READMEs, lookalike profiles, and malicious ZIP files to trick users into downloading and executing SmartLoader malware. The campaign has also recorded more than 14 million downloads across GitHub Release assets from around 200 campaign repositories. Once executed, SmartLoader can establish persistence and deliver secondary malware such as StealC, an information stealer capable of harvesting sensitive information from compromised systems. The primary reason behind the campaign is to exploit the growing popularity of AI tools, agentic workflows, and MCP servers. Attackers created fake repositories that appeared to provide useful integrations for services such as Gmail, WhatsApp, Databricks, Jenkins, and Docker. The campaign introduced a technique called AgentBaiting, where AI agents searching for skills or MCP servers may independently discover malicious repositories and follow instructions contained in their READMEs. This creates a new attack path in which malicious content can influence AI agents without requiring attackers to directly send a malicious link to a human user. The attack begins when a user or AI agent discovers a fraudulent repository and follows its installation instructions. A malicious ZIP archive is then downloaded, triggering a LuaJIT-based loader chain that executes an obfuscated Lua script and installs SmartLoader. The malware can subsequently deploy StealC to steal sensitive data. To reduce the risk, organizations should maintain an approved catalog of AI skills and MCP servers, verify publishers and repositories, sandbox new tools before deployment, and monitor AI agent activities and installation workflows for suspicious behavior.