npm security reports, with evidence.

Source inspection confirms a bundled all-sites credential collector that uploads accepted login secrets to a remote default endpoint. This behavior is materially unrelated to the advertis...

Direct source inspection confirms automatic install-time system mutation and runtime disclosure of WhatsApp content to a hard-coded external Telegram recipient. These behaviors establish...

This is a concrete credential-collection and transmission chain to a hard-coded raw-IP endpoint with TLS verification disabled. The lack of an npm lifecycle hook limits installation risk...

This package performs automatic postinstall mutation of foreign AI-agent control surfaces and installs additional user-level dependencies. That meets the install-hook abuse blocking crite...

This is a concrete remote payload-execution chain that forwards sensitive values to remotely selected code. The lack of an install hook does not mitigate the explicit runtime attack surface.

The declared runtime entrypoint directly invokes an obfuscated remote-binary downloader and process launcher. The absence of a lifecycle script limits the install-time trigger but does no...

This is a concrete automatic install-time payload chain: obfuscated code downloads and detaches a remote executable. The behavior is unrelated to the package's stated API-wrapper role.

The package contains a concrete, default-enabled runtime data-exfiltration path to embedded external webhook receivers. There is no install hook, but that does not remove the runtime disc...

The inspected source establishes automatic install-time mutation of consumer lifecycle and agent settings, including persistence through package.json. This meets the blocking boundary for...

The package embeds active third-party webhook destinations and automatically sends serialized runtime records to them during normal module use. Although this is not install-time behavior,...

This package performs an unconsented postinstall mutation of a foreign global AI-agent control surface. Its runtime server also exposes unauthenticated shell command forwarding on a non-l...

The source establishes an automatic postinstall chain that mutates foreign consumer projects and persists itself in their lifecycle configuration. This meets the blocking policy for uncon...

The package performs automatic postinstall mutation of a user-level Claude SessionStart control surface, creating recurring agent-session execution. This meets the install-hook abuse bloc...

The package performs unconsented postinstall mutation of a user-level AI-agent control surface and establishes future command execution. This meets the install-hook abuse blocking policy...

The package contains a concrete automatic postinstall chain that changes global AI-agent configuration and installs a command hook for future sessions. No exfiltration was found, but the...

The postinstall hook mutates both global agent skill directories and installs user-level dependencies without a user-invoked setup action. This meets the install-hook abuse boundary even...

The lifecycle hook automatically installs and prunes agent skills in both Codex and Claude user configuration trees. This meets the policy threshold for malicious install-hook abuse despi...

This is an unconsented postinstall mutation of a broad AI-agent control surface, with package-authored instructions registered for future use. It meets the install-control-surface block p...

This is a concrete remotely controlled payload-deployment channel into OpenClaw skill directories, activated automatically at gateway startup. The lack of an npm lifecycle hook does not n...

This package contains an automatic, deliberately obfuscated installer that conditionally downloads and executes an external binary. That is a concrete malicious install-hook chain unrelat...

The package implements an import-triggered detached remote-code loader rather than a bounded client feature. Its remote payload receives Node module-loading capability, establishing concr...