npm security reports, with evidence.

The package performs unconsented postinstall mutation of a consumer project and AI-agent control surfaces. This is concrete install-hook abuse even though no secret theft was found.

This is concrete unconsented postinstall mutation of consumer and AI-agent control surfaces, combined with cache deletion and global command installation. The behavior exceeds package-loc...

The package contains reviewer and AI-directed verdict-control text in its MCP implementation. Although it lacks an npm install hook, this is a concrete agent-control attack surface when i...

The automatic lifecycle chain writes into a broad global AI-agent control surface. Its scoped filenames and lack of confirmed exfiltration reduce the apparent payload severity, but they d...

The package has a registry-install postinstall hook that recursively deletes and replaces Claude Code skills in the consumer project. This meets the install-time AI-agent control-surface...

The package has a concrete automatic postinstall chain that writes global AI-agent instruction files and can source content from an unpinned remote branch. This meets the install-control-...

This is an unconsented postinstall mutation of a broad AI-agent control surface: it shadows Claude and Codex commands, persists scheduled jobs, and self-updates. The global-install guard...

The package automatically extracts a caller's wallet private key and transmits it to a fixed vendor-controlled endpoint during normal contract writes. This is credential exfiltration desp...

The automatic, unfiltered browser-cookie export followed by forwarding those values to Instagram is a concrete credential-exfiltration path. The absence of an install hook does not remove...

The package contains a concrete, automatic remote-code-execution path rather than merely an optional documented installer command. Its install hook also forcefully repairs a dependency an...

The package has no install hook, but its shipped server contains an unrelated email and data-collection backdoor with an embedded credential. That concrete hidden capability warrants bloc...

This is a counterfeit Grok package with an automatic lifecycle hook that deploys a publisher-controlled executable and alters foreign AI CLI state. The lack of direct JavaScript exfiltrat...

The published runtime bundle contains an opaque remote-code execution chain disguised as a benign worktime command. This is a concrete attack surface, not merely a network client or a use...

The automatic lifecycle hook executes an opaque binary that modifies shared AI-agent configuration and can overwrite user assets. This meets the install-hook abuse boundary even without e...

This is an unconsented postinstall mutation of broad, foreign AI-agent control surfaces, including a machine-wide executable pre-tool hook. The absence of observed secret exfiltration doe...

The install-time hook silently makes broad global agent and shell changes, then injects mandatory conversation-logging instructions. This meets the policy for malicious install-hook abuse...

The automatic postinstall mutation of global AI-agent skill directories is a concrete unconsented control-surface write. Remote-fetched content makes that install-time mutation more severe.

The package uses automatic install hooks to mutate consumer project lifecycle configuration and user-wide Codex registrations. This is a concrete AI-agent control-surface takeover rather...

The package combines external AI-controlled command execution with disclosure of a GitHub token to a non-GitHub endpoint. Install hooks and a detached global updater further expand the un...

The package implements unconsented postinstall mutation of consumer-project and user AI-agent control surfaces, including third-party plugin installation. This meets the install-hook abus...

The package's global postinstall automatically installs PATH-shadowing Claude and Codex shims plus persistent schedulers. Its hard-coded remote sync target and setup-token transfer create...