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, 13:52 UTC. Ordered by latest scan.
The automatic postinstall mutates a foreign Claude-related AI-agent surface and deletes files from ~/.claude; this meets the block boundary. The unchecked executable download adds further...
Source confirms a deliberate GitLens license-bypass and foreign extension overwrite capability. This is concrete malicious behavior despite not being activated by postinstall.
Concrete postinstall code mutates global Claude and Codex skill roots and installs instructions that accept injected work. This meets the firewall block threshold regardless of the absenc...
The package performs a concrete, destructive system-wide executable replacement during postinstall. Its transparency reduces deception risk but does not remove the harmful install-time at...
The package contains concrete, automatic, unconsented mutation of outgoing user media metadata. Although the lifecycle hook is benign and no host credential theft was found, the hidden at...