npm security reports, with evidence.

This is unconsented postinstall mutation of broad foreign AI-agent control surfaces, including destructive replacement. The behavior meets the install-hook abuse blocking policy.

This is an unverified remote-code launcher over HTTP with automatic update behavior and an external-package fallback. Although it lacks an install hook, the explicit CLI invocation create...

The package performs automatic postinstall AI-agent hook and configuration changes through an opaque executable. This is a concrete unconsented control-surface modification, not an explic...

The package performs automatic install-time network retrieval and execution of an opaque binary, then automatically installs agent skills. This creates a concrete unconsented AI-agent con...

The package contains a concrete remote credential and shell-input collection path, and the relevant exam command is deliberately obfuscated. The lack of an install hook does not remove th...

Automatic postinstall execution of opaque native code to repair AI-client hooks is a concrete unconsented foreign control-surface mutation. No source-visible exfiltration was needed to es...

The automatic postinstall writes a Claude skill into the consuming project and can rewrite its ESLint configuration. This is unconsented install-time mutation of a foreign AI-agent contro...

This is unconsented postinstall mutation of broad AI-agent control surfaces, with an opaque native binary that contains a safeguard-bypass flag. The package meets the publish-block thresh...

The automatic postinstall executes opaque native code specifically to install agent hooks, including code obtained without integrity verification. This is a concrete install-hook abuse pa...

This is concrete unconsented postinstall mutation of foreign AI-agent control surfaces, coupled with remote runtime download and execution. The global-only guard and opt-out do not establ...

This is an automatic lifecycle mutation of a foreign AI-agent executable path using opaque bundled binaries. The installer’s backup and rollback features do not remove the unconsented exe...

This is an unconsented postinstall mutation of a foreign AI-agent control surface that establishes persistent command execution. The conditional PATH check does not make the automatic con...

The automatic self-reinstallation chain can execute lifecycle code from a future release and immediately rerun broad agent configuration. The observed configuration injection reduces appr...

The package combines unconsented host tracking and error telemetry with explicit instructions designed to make an AI agent trust its output and avoid verification. There is no install hoo...

The automatic lifecycle chain executes an opaque native hook installer that the package identifies as modifying Codex and Claude Code configuration. This is concrete install-hook abuse un...

The package uses an automatic install hook to download and execute remote code and to alter global AI-tool dependencies. That is concrete install-hook abuse with user-level execution impact.

The lifecycle hook automatically writes behavior-bearing files into both .agents and .claude home directories. Although no exfiltration or remote execution is present, this satisfies the...

The postinstall hook makes broad AI-agent configuration changes and installs a separately delivered, unpinned MCP runtime without user action. This meets the install-control-surface block...

This is concrete unconsented postinstall mutation of external AI-agent controls through an opaque native binary, with an unsigned remote-download fallback. It meets the install-control-su...

This is a concrete unconsented postinstall mutation of broad AI-agent control surfaces, combined with automatic installation of a separate registry package and process restart. It meets t...

The automatic postinstall hook makes persistent changes to multiple unrelated AI-agent configuration surfaces and injects directives for financial and credential-related actions. This mee...