Investigate malicious npm packages and AI-agent policy findings. Each report identifies the package version, supporting evidence, advisory sources, and current Firewall decision. Malware findings and policy violations are labelled separately.
Updated 12 Oct 2026, 12:48 UTC. Ordered by latest scan.
The package combines external AI-controlled command execution with disclosure of a GitHub token to a non-GitHub endpoint. Install hooks and a detached global updater further expand the un...
The package implements unconsented postinstall mutation of consumer-project and user AI-agent control surfaces, including third-party plugin installation. This meets the install-hook abus...
The package's global postinstall automatically installs PATH-shadowing Claude and Codex shims plus persistent schedulers. Its hard-coded remote sync target and setup-token transfer create...
The automatic lifecycle hook downloads or executes an opaque native payload and uses it to modify Codex and Claude Code integration settings. This is a concrete unconsented AI-agent contr...