Malicious agent-framework releases on PyPI targeted credentials
Malicious versions of Microsoft's agent-framework Python package were published to PyPI and targeted credentials in developer environments and CI pipelines.
According to SecurityWeek, two malicious versions of Microsoft’s Agent Framework Python package were distributed through PyPI (the Python Package Index) on November 17, 2026. The report identified them as versions 1.26.1 and 1.26.2 of the agent-framework package. Both have since been quarantined on PyPI.
What is affected
The package is agent-framework, which the report says has approximately 190,000 downloads per month. The malicious releases contained only built packages, unlike legitimate releases that included both built and source distributions. The method used to publish them remains unknown. The report said SafeDep identified a PyPI maintainer account named vibiret, created on the day of the attack, with publishing permissions for the affected package and five related projects.
Why it matters
The malware runs whenever the Python interpreter starts, even if the application does not explicitly import the compromised package. Its targets include secrets in environment variables and configuration files, credentials of other running processes, and Azure managed identity and Azure DevOps authentication tokens. The second malicious release expanded credential-stealing functionality to GitHub Actions. Microsoft acknowledged the incident, and an Agent Framework maintainer confirmed the compromised releases were removed from PyPI and that an investigation was underway.
What to do
The report says version 1.26.0 remains the latest clean release. Check developer environments, CI/CD (continuous integration and continuous delivery) pipelines and container images for the compromised versions, then replace affected installations with verified packages. Rotate potentially exposed credentials, investigate suspicious cloud activity, and rebuild affected environments.
Sources
Related reading
3 postsDiscussion
Spotted an error? Send corrections to [email protected] .