Skip to content
NewA calmer, faster ZeroHour

Malicious Agent Framework PyPI releases stole cloud credentials

Original title: Microsoft’s Agent Framework PyPI package compromised to steal credentials

BleepingComputerby Bill Toulas ()2 min read
Part of a story covered by 9 sourcesPyPI quarantines malicious Microsoft agent-framework 1.26.1 and 1.26.2Merged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

Versions 1.26.1 and 1.26.2 ran a stealer at Python startup on November 17; the safe release named is 1.26.0.

Attackers published two malicious versions of Microsoft's official Agent Framework package on PyPI to steal cloud credentials and other secrets. Versions 1.26.1 and 1.26.2, published November 17, 2026, executed during normal Python startup even when the package was not imported, while depending on legitimate agent-framework-core 1.26.0. They collected environment variables, configuration files, process data, and cloud tokens, then sent them to a webhook.site address. PyPI quarantined both releases, but the publishing-access path remains unknown.

Unconfirmed: How the attacker gained publishing access to Microsoft's official PyPI project is unknown, and Microsoft had not responded.

Key facts
  • The package has roughly 200,000 monthly downloads on PyPI.
  • Targets included Azure, AWS, GitHub, Docker, Hugging Face, and OpenAI secrets.
  • Version 1.26.2 added GitHub Actions OIDC requests and runner-directory searches.
  • The second malicious release appeared about 2.5 minutes after the first quarantine.
  • Both releases sent collected data to the same webhook.site endpoint.
Security assessment · AI, from the article

Impact: Theft of cloud credentials, tokens, and other developer secrets

ProductAffectedFixed in
Microsoft Agent Framework1.26.1 and 1.26.21.26.0
What to do
  1. 1.Isolate affected environments and rebuild from the safe 1.26.0 release.
  2. 2.Rotate credentials reachable from any compromised environment.
  3. 3.Search egress logs for the named webhook.site path.
Analyze this story with Vani AIImpact, who is affected and what defenders should do, in a new chat

Indicators of compromise

auto-extracted · verify before use · export allAll
TypeIndicatorContext
domainwebhook.sitected data to the same attacker-controlled endpoint, https://webhook[.]site/d381fd80-ec62-42ad-8f5c-d1b15103c9bb . Collecting data fr

From the original article

2 min read at bleepingcomputer.com

Hackers compromised Microsoft's official Agent Framework Python package on PyPI, publishing two malicious versions designed to steal cloud credentials and other sensitive information from developers' systems. [...]

BleepingComputer

By Bill Toulas · bleepingcomputer.com

Read the full article at BleepingComputer

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.