npm AI-agent policy reports, with evidence.

The automatic postinstall hook modifies the consuming repository's Claude Code instruction surface. This meets the policy threshold for malicious install-hook abuse even though no network...

The automatic global installation of a separate AI agent is concrete install-time behavior, not required for package extraction or ordinary dependency setup. Its later default automatic-a...

The package has a concrete automatic install-time chain that force-upgrades registered foreign projects and mutates broad AI-agent control surfaces. This meets the install-control-surface...

The automatic global-install hook performs remote shell execution and unconsented Codex marketplace/plugin mutation. Its guards limit scope but do not provide user consent for this broad...

This is an automatic postinstall mutation of a foreign AI-agent control surface, not an explicit user-command setup flow. The hook installs behavioral instructions that influence future a...

The automatic lifecycle hook performs broad external AI-agent and shell configuration changes. That concrete install-time control-surface mutation meets the publish-block policy despite n...

The automatic lifecycle hook performs unconsented global installation of a separate AI-agent CLI and package manager, plus host cleanup. This is concrete install-time control-surface muta...

Source inspection confirms an install-time chain that mutates a consumer package.json and AI-agent plugin state, rather than merely providing user-invoked tooling. The dev-only self-refer...

This is an unconsented post-install mutation of broad shell and AI-agent control surfaces. The persistent hook changes user-wide Claude permissions from project-controlled input, so it me...

Source inspection confirms an automatic postinstall that performs broad consumer-project and AI-agent configuration changes, including plugin installation and detached follow-up execution...

The package performs unconsented install-time mutation of a consumer AI-agent control surface and adds third-party agent plugins. This is concrete install-hook abuse rather than a user-in...

This is an unconsented postinstall mutation of broad AI-agent and repository control surfaces in the consuming project. The lack of observed network exfiltration does not neutralize the a...

The package performs automatic install-time changes to consumer project and AI-agent configuration rather than limiting this behavior to an explicit user command. No direct credential the...

This is a concrete, automatic remote agent-control channel rather than a user-invoked updater: it connects by default and accepts remote archives for skill replacement. The lack of comman...

This is an unconsented install-time mutation of a broad AI-agent control surface, followed by persistent promotion of project permissions into global settings. The behavior meets the inst...

The package has a concrete postinstall chain that automatically writes bundled content into shared Claude and Codex agent control surfaces. This meets the install-hook abuse block conditi...

The automatic global postinstall mutation of multiple foreign or shared AI-agent skill surfaces is a concrete install-hook abuse. No exfiltration is needed for this policy outcome.

This package performs unconsented postinstall mutation of OpenCode and consumer-project dependency control files, then reinforces it at runtime with automatic global skill installation an...

The automatic postinstall hook broadly mutates global Claude and Codex skill directories and removes selected command links. This meets the install-time foreign AI-agent control-surface a...

The package performs unconsented postinstall mutation of a foreign AI-agent control surface and broad deletion in that control surface. Those actions create a concrete harmful install-tim...

The package has a concrete automatic postinstall path that mutates a foreign project and AI-agent plugin configuration, including a detached persistence-style reconciliation process. No s...