More than 70 extensions that were published to the Open VSX marketplace in April are likely sleeper extensions linked to the GlassWorm malware, Socket reports.
GlassWorm appeared in the Open VSX registry in October 2025 in a dozen extensions that were likely downloaded thousands of times. It used Unicode variation selectors to visually hide its code, and relied on the Solana blockchain for command-and-control (C&C) infrastructure.
Designed to steal GitHub, Git, and NPM credentials, sensitive information, and cryptocurrency, GlassWorm spread to other open source software ecosystems in November and re-emerged in January and again in March, when it compromised over 150 repositories.
Now, Socket says it has identified 73 suspicious extensions that are clones of popular extensions on the Open VSX marketplace. They were published by newly created GitHub accounts that have one or two public repositories named with an eight-character string.
All these extensions are likely sleepers designed to deploy malware on the users’ machines through future updates, and at least six of them have been activated.
“This count may change as new updates continue to appear, but the pattern is consistent with earlier GlassWorm waves: cloned or impersonating extensions are first published without an obvious payload, then later updated to deliver malware through the normal extension update path,” Socket says.
The extensions feature a clear impersonation pattern, where they mirror the legitimate listings of the cloned extensions, including icons, naming, and description, but under a different publisher and unique identifier.
“This is the core social engineering pattern behind the latest GlassWorm cluster: cloned listings create enough visual trust to attract installs before any malware is introduced,” Socket notes.
The malware delivery method implemented by these extensions is a combination of previously observed mechanisms: some rely on bundled native binaries, including components from previous GlassWorm attacks, while others retrieve the payload from a remote location.
“The extension’s source code alone no longer reflects the behavior that ultimately runs. By shifting critical logic outside of what tools typically scan, and spreading it across multiple delivery mechanisms, the threat actor increases the likelihood of evading detection,” Socket notes.
Related: Bitwarden NPM Package Hit in Supply Chain Attack
Related: Are SBOMs Failing? Supply Chain Attacks Rise as Security Teams Struggle With SBOM Data
Related: ‘By Design’ Flaw in MCP Could Enable Widespread AI Supply Chain Attacks
Related: European Commission Confirms Data Breach Linked to Trivy Supply Chain Attack
