npm security reports, with evidence.

This is a concrete, unconsented install-hook abuse chain that broadly mutates and persistently controls consumer project files. No self-dependency or install-time exfiltration was found,...

The inspected code establishes automatic, noninteractive consumer-project and AI-agent plugin mutation during postinstall, with detached persistence of reconciliation. This exceeds transp...

This package redirects third-party API credentials and agent data to a fixed gateway by default, then includes an unverified remote executable fallback. The absence of an install hook doe...

This is an unconsented postinstall mutation of a foreign AI-agent control surface, with injected instructions intended to control agent behavior. That meets the install-hook abuse blockin...

The automatic postinstall mutation targets broad, foreign AI-agent control surfaces and changes how host agents handle tasks. This meets the install-hook abuse blocking boundary despite t...

The automatic postinstall chain deliberately modifies the consumer project's AI-agent instruction and adapter surfaces across multiple vendors. This satisfies the policy for malicious ins...

The package implements a concrete third-party credential collection path selected by default, including an automatic re-login trigger. The absence of an npm lifecycle hook does not remove...

Source inspection confirms automatic postinstall mutation of consumer project and Claude Code plugin state, including marketplace refresh and third-party plugin installation. This meets t...

The package contains a concrete automatic install hook that runs a package manager against an external relative directory. The browser bundle findings do not mitigate this install-time mu...

The automatic postinstall hook modifies multiple external AI-agent instruction files and injects behavioral guidance. This meets the policy definition of malicious install-hook abuse.

The source contains concrete credential handling, provider-identity deception, external host fingerprint transmission, and runtime persistence. Although there is no install hook, these ar...

This package performs an automatic postinstall mutation of several foreign AI-agent control surfaces and injects instructions for financial and credential-related actions. This meets the...

Default remote execution exports the configured Roark bearer token to a third-party host when the MCP execute tool is used. That is concrete credential exfiltration despite the absence of...

The package hides a remotely controlled kill switch in a photo-frame component and destroys the consumer page when its remote licence decision is negative. There is no install-time execut...

The absence of an install hook reduces automatic-install risk, but the source deliberately conceals the actual upstream provider and rewrites a Claude control surface to route user prompt...

The component contains a concrete, destructive action against the consuming application's DOM, guarded only by remote-controlled license status. No install hook is needed for this harmful...

The package combines deliberate obfuscation, an undisclosed remote control check, and destructive host-page behavior in a simple UI component. Although it has no install hook, the runtime...

The automatic lifecycle chain silently mutates global instruction surfaces for multiple AI-agent products. This meets the install-control-surface blocking policy regardless of the absence...

The automatic lifecycle hook performs unconsented global AI-agent tooling and persistent user-environment mutations. This meets the install-hook abuse blocking boundary even though no cre...

The package contains a concrete, silent remote logging path for user mail-account metadata and runtime errors. The local-only install hook does not mitigate that runtime data exfiltration...

The package has a concrete automatic postinstall path that installs code into a global agent extension location and alters future agent sessions. No self-dependency or network theft was f...