@miracleplus/mplus@0.6.11

High-risk behavior combination matched malicious policy.

npm installation, through the postinstall lifecycle hook.

Package-controlled instructions can become available to installed AI agents without an explicit setup command.

Recursive copying of bundled agent skills and commands into home-directory agent control surfaces.

On installation, npm invokes the package's postinstall script. That script creates ~/.agents/skills and copies every bundled skill directory there. If Claude or Codex directories already exist, it also copies package-controlled skills or commands into their user-level control locations. The operation is automatic and failure is suppressed, so no explicit agent-setup command or confirmation is required.

This is an unconsented postinstall mutation of broad and foreign AI-agent control surfaces. The lifecycle hook and recursive overwrite behavior establish concrete install-hook abuse despite no observed network activity in that hook.