AI Threat Watch

Case mini-shai-hulud-ai-packagesAI supply chainWorldwide

Mini Shai-Hulud worm hits TanStack, Mistral AI and Guardrails AI packages

TeamPCP's worm hijacked release pipelines to publish 170+ malicious npm/PyPI versions, stealing cloud, CI and AI-tool credentials and planting persistence in Claude Code and VS Code.

CriticalAttack on AIStatus: ContainedVerified case file

Open in the interactive mapCite this case

Reported
May 12, 2026
Location
Worldwide
Target
npm/PyPI AI developer packages
Severity
Critical
Triggered by
TeamPCP
Actor type
Criminal
Reported loss
None reported
CVE
None

What happened

TeamPCP's worm hijacked release pipelines to publish 170+ malicious npm/PyPI versions, stealing cloud, CI and AI-tool credentials and planting persistence in Claude Code and VS Code.

Impact: 170+ malicious versions across 19 npm namespaces plus PyPI, including packages with 1.3M weekly downloads; stole GitHub, AWS, Vault, Kubernetes, npm and AI-tool credentials and planted persistence hooks in Claude Code and VS Code

Who triggered it

Actor: TeamPCP

Attribution: Attributed to TeamPCP by Wiz, Socket, Expel and Phoenix Security based on tooling overlap with earlier campaigns

Motive: Mass credential theft (cloud, CI, npm, AI-tool tokens) for resale and further compromise

How it happened

Attack vector: Fork pull request against TanStack router poisoned the GitHub Actions cache, extracted an OIDC publish token and published trojanized packages; worm then reused stolen npm tokens to spread

Root cause: GitHub Actions cache collision between fork and upstream PRs, overly broad token permissions and build-plus-publish in one job, defeating 2FA and trusted publishing

Timeline

  1. Attacker fork of TanStack/router created; orphaned commit plants malicious workflow
  2. Malicious TanStack, OpenSearch, Mistral AI and Guardrails AI versions published to npm and PyPI
  3. Socket, Wiz and Expel publish analyses; maintainers yank versions and rotate tokens
  4. Related fork-triggered compromise hits openapi-react-query-codegen

Response

Compromised versions removed; TanStack published a post-mortem; vendors issued IOCs and cleanup guidance for IDE persistence hooks; affected users told to rotate all secrets

Current status: Contained

Sources (open in a new tab)

Cite this case

Mini Shai-Hulud worm hits TanStack, Mistral AI and Guardrails AI packages. AI Threat Watch, case mini-shai-hulud-ai-packages, May 12, 2026. https://aithreat.io/cases/mini-shai-hulud-ai-packages/

Permanent link: https://aithreat.io/cases/mini-shai-hulud-ai-packages/

This case file in other languages: EnglishFrançaisEspañolالعربية中文