What happened
Two compromised releases of agent-framework, the Python package behind Microsoft's toolkit for AI agents and multi-agent applications, reached PyPI on November 17 carrying credential-harvesting code, SecurityWeek reported. Versions 1.26.1 and 1.26.2 landed about 33 minutes apart and are now quarantined, leaving 1.26.0 as the latest clean build. A project maintainer confirmed on GitHub that they were removed and an investigation had begun.
Analyses by OX Security, Wiz, Endor Labs and SafeDep found the malware abuses a Python startup mechanism, so it executes each time the interpreter launches, whether or not the package is imported. It hunts for secrets in environment variables, configuration files and other running processes, and tries to obtain Azure managed identity and Azure DevOps tokens. The second release extended its reach to GitHub Actions.
How the packages were published remains unclear. Genuine releases ship with both built and source distributions, but the rogue ones had only built files, and Wiz found no matching code in Microsoft's main repository. SafeDep pointed to a PyPI maintainer account, vibiret, created that day with publishing rights over the package and five related projects, though its role has not been confirmed.
Context
Microsoft Agent Framework is meant for developing AI agents and multi-agent systems. Organizations are being advised to scan developer machines, CI/CD pipelines and container images, swap in verified packages, rotate exposed credentials and rebuild affected environments. The framework's GitHub repository was inaccessible at the time of reporting, as were several other Microsoft repositories including ONNX Runtime projects, with no confirmed link.
ब्रीफ़िंग
टेक का पूरा सप्ताह, एक ईमेल में।
हर सप्ताह वे रिपोर्ट्स जो आपके समय के लायक हैं। कभी भी अनसब्सक्राइब करें।
