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, 20:11 UTC. Ordered by latest scan.
Concrete install-time mutation of both Claude Code and Codex global MCP configuration meets the block policy for unconsented foreign/broad AI-agent control-surface writes.
The source confirms a concrete npm postinstall chain that writes and merges package-controlled content into broad external AI-agent control surfaces. This meets the blocking policy regard...
Direct manifest inspection confirms an install-time command that mutates the consuming project by installing a separate package. No benign package-aligned setup purpose is shown.
The lifecycle hook performs an unconsented install-time mutation outside the package directory. The missing declared entrypoint does not mitigate that concrete behavior.
Direct source inspection confirms an install-time mutation of the consuming project's manifest to add an unrelated dependency. This is concrete malicious supply-chain behavior.