Miasma Worm Steals Cloud Keys via 32 Red Hat npm Packages
Attackers exploited a compromised Red Hat GitHub account to publish 32 malicious npm packages containing the Miasma credential-stealing worm.
On June 1, 2026, attackers published 32 malicious npm packages under the official @redhat-cloud-services namespace, delivering a credential-stealing worm known as Miasma. The supply chain attack executed within a single 72-second window, leveraging compromised GitHub Actions OIDC tokens to bypass traditional npm authentication. For developers maintaining build pipelines, the incident exposes the risks of trusted publishing mechanisms when the underlying identity provider is breached.
The Miasma Payload
Miasma is an evolved variant of the Mini Shai-Hulud malware. Both are attributed to the TeamPCP threat group, which recently executed a similar supply chain attack against Python packages. Miasma executes automatically via a preinstall hook in the package’s package.json file. This design triggers the malicious payload immediately upon running the standard npm install command, regardless of whether the developer actually imports the library into their application code.
The malware heavily relies on JavaScript obfuscation, combining nested eval() statements and ROT-based decoding to hide its primary objectives from static analysis tools. Once executed, Miasma scans the local development environment for high-value secrets.
| Capability | Mini Shai-Hulud | Miasma |
|---|---|---|
| Target Cloud Keys | AWS, GCP | AWS, GCP, Azure |
| Local Secrets | GitHub, Kubernetes | GitHub, Kubernetes, Vault, 1Password |
| Identity Collection | None | Dedicated GCP and Azure modules |
| Propagation Method | Static | Wormable via stolen write tokens |
Unlike earlier iterations, Miasma actively maps cloud environments using dedicated GCP and Azure identity collectors. It also functions as a self-propagating worm. The script attempts to use any stolen GitHub or npm tokens to republish itself to other repositories where the compromised user has write access.
OIDC and Trusted Publishing Risks
The affected frontend libraries and UI components primarily support Red Hat’s Hybrid Cloud Console. They collectively average 80,000 to 117,000 weekly downloads. Red Hat confirmed the breach in Security Advisory RHSB-2026-006, noting that the compromised packages only affected internal development tooling rather than customer production environments.
Security researchers identified the root cause as a compromised Red Hat employee GitHub account. The attackers used GitHub Actions OpenID Connect (OIDC) tokens to push the malicious updates to the npm registry. This trusted publishing architecture improves security by eliminating long-lived npm tokens, but it completely relies on the integrity of the initial GitHub identity. When that identity is compromised, attackers inherit the pipeline’s full publishing authority. This attack vector mirrors recent incidents where OIDC theft enabled signed malware in other major open-source ecosystems.
Red Hat revoked the affected npm versions and rotated the compromised infrastructure credentials. They also removed the malicious code from the canonical repositories.
If you installed any @redhat-cloud-services packages between June 1 and June 2, 2026, assume your local environment is compromised. You must immediately rotate all local secrets, including SSH keys, cloud provider tokens, and registry credentials, before investigating further.
Get Insanely Good at AI
The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.
Keep Reading
How to Secure Claude API Workloads With Identity Federation
You will learn how to configure Workload Identity Federation to authenticate non-human Claude API requests and eliminate static access keys.
PyPI and npm Purge 73 Fake Azure Packages Targeting AI Agents
Security researchers discovered 73 malicious PyPI and npm packages mimicking Microsoft Azure libraries to install credential stealers on AI coding agents.
PyTorch Lightning 2.6.2 Drops Self-Spreading Credential Stealer
Threat actors hijacked the PyPI credentials for PyTorch Lightning to publish two malicious versions containing a self-propagating credential stealer.
Context Bombing Drops AI Attacker Success Rates to 5%
Tracebit researchers developed context bombing to plant forbidden prompts alongside cloud secrets, triggering model guardrails to neutralize AI attackers.
57% Bot Traffic Triggers Cloudflare Precursor Release
Cloudflare's new Precursor engine combats the rise of agentic web automation by replacing point-in-time challenges with continuous client-side telemetry.