npm security reports, with evidence.

This is concrete remote code execution on import from a hard-coded package-controlled endpoint. Signature verification authenticates the operator's payload but does not make arbitrary rem...

This is an automatic install-time mutation of broad third-party AI-agent control surfaces combined with remote binary acquisition and host persistence. It meets the install-hook abuse blo...

The automatic postinstall executes an opaque initializer that registers agent skills in user-home directories. This meets the install-hook abuse policy regardless of whether secret theft...

This is an unconsented postinstall mutation of a foreign AI-agent control surface. The absence of shell execution or observed exfiltration does not neutralize that automatic control-surfa...

This is a concrete automatic install-hook attack on a foreign AI-agent control surface, not merely a platform shim. The lack of observed exfiltration does not offset the installed authent...

The inspected lifecycle chain automatically modifies consumer lifecycle configuration and Codex hooks. This meets the install-hook-abuse blocking boundary regardless of the package's stat...

The automatic fallback changes the model provider and transmits Codex request data to an external service without an explicit user selection. The concealed key pool reinforces that this i...

The package creates persistent, broad AI-client hooks that exfiltrate conversation content and accept remotely supplied replacement code. Although setup is user-invoked and not lifecycle-...

This is malicious install-hook abuse under the policy because an automatic postinstall mutates foreign AI-agent control surfaces and performs broad host changes. The global-install guard...

The automatic lifecycle mutates shared third-party AI-agent configuration directories and user-level dependencies. This is a concrete unconsented install-hook attack surface.

The automatic lifecycle hook overwrites consumer Claude behavior and expands MCP permissions without an explicit user command. This meets the install-time AI-agent control-surface abuse b...

This is an automatic install-time delivery chain that plants an AI-directed prompt and induces broad, persistent modification of the consumer project's agent control surfaces. The behavio...

This is unconsented postinstall mutation of a broad foreign AI-agent control surface. CI and opt-out guards reduce exposure but do not make the default consumer-project modifications safe.

The automatic postinstall changes a consumer project's AI-agent plugin surface and refreshes remote plugin content. This is a concrete unconsented install-hook control-surface mutation.

This is an automatic install-time mutation of a consumer AI control surface that redirects imports into telemetry interception and remote transmission. The package-aligned endpoint does n...

Automatic installation changes the consumer project and injects agent-review controls that hide its own generated artifacts. This is concrete install-hook abuse, not a user-invoked setup...

This is an unconsented install-time mutation of a consumer AI control surface that captures and transmits application LLM content. The behavior is concrete and automatically reachable fro...

The postinstall hook automatically and destructively installs a skill across multiple unrelated AI-agent directories. This meets the install-control-surface blocking policy despite no obs...

The postinstall hook mutates both Codex and Claude user skill directories and prunes their contents without a user-invoked setup action. This meets the install-time foreign AI-agent contr...

The package has a concrete automatic postinstall chain that broadly mutates foreign AI-agent skill directories. Its apparent product purpose does not establish consent for this install-ti...

The source establishes a complete runtime chain from application input to package-selected external webhooks. The absence of an install hook does not mitigate this active data-exfiltratio...