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 17 Oct 2026, 01:34 UTC. Ordered by latest scan.
This is a concrete npm postinstall mutation of broad, foreign AI-agent configuration surfaces, including an executable Codex notify hook. The inspected runtime LLM networking is not neede...
This is concrete, unconsented install-time metadata exfiltration, regardless of its claimed research purpose. The DNS encoding and redundant callback channels make the behavior an active...
Concrete postinstall writes target global Claude and detected Codex control surfaces rather than a package-owned directory. This meets the firewall block policy for unconsented install-ti...
This is concrete install-time credential exfiltration plus remote payload placement, unrelated to TOTP functionality. The benign API is a cover and also activates the malicious runner.