npm AI-agent policy reports, with evidence.

The install hook performs unconsented mutation of broad, foreign AI-agent control surfaces and deletes existing target directories before replacing them. This meets the blocking policy ev...

The package performs unconsented postinstall installation of a foreign AI-agent CLI plus persistent shell or registry mutation. This is a concrete install-time control-surface change, not...

The package performs unconsented privileged postinstall creation of a persistent TLS interception service and changes AI request content. Its scoped host and absence of exfiltration reduc...

This is deliberately disguised, persistent proctor-evasion software that captures exam content and automates AI answers. The harmless install hook does not mitigate the concrete malicious...

The package performs unconsented postinstall mutation of multiple foreign AI-agent control surfaces. This meets the firewall block policy despite the absence of confirmed exfiltration or...

This package performs unconsented postinstall mutation of broad, foreign AI-agent control surfaces across registered projects. That meets the blocking boundary even though no data exfiltr...

Source inspection confirms an automatic foreign agent configuration write and instructions for unapproved outbound actions. The harmless-looking telemetry hook does not mitigate that conc...

Source inspection confirms concrete, unconsented postinstall writes to global Claude Code and Codex MCP configuration. This meets the block policy for foreign and broad AI-agent control-s...

This is a concrete unconsented postinstall mutation of foreign AI-agent control surfaces. It meets the firewall block policy even though the inspected hook does not itself exfiltrate data.

This is an unconsented postinstall mutation of foreign AI-agent control surfaces. Documentation does not remove the install-time control-surface risk.

The lifecycle hook performs forced upgrades of registered external projects, including agent instruction and skill locations, rather than limiting setup to the package directory. This mee...

Source inspection confirms an unconsented postinstall writes to Herdr's plugin and configuration paths, then installs a detached clipboard-handling process. This meets the block policy fo...

The source implements a concrete unconsented global postinstall chain that changes third-party AI-client control surfaces and installs remote native code. Product-aligned endpoints do not...

The package performs an unconsented global-install postinstall mutation of Claude’s skill directory and replaces conflicting symlinks. Its explicit CLI network and upload behavior does no...

The package's global postinstall mutates several foreign AI-agent skill directories automatically. That concrete lifecycle behavior meets the firewall block boundary even though its searc...

The package performs unconsented postinstall mutation of broad, foreign AI-agent skill directories. This meets the firewall block rule even though no credential theft was found.

The automatic postinstall mutation of broad, foreign AI-agent skill directories is a concrete unconsented control-surface change. No credential theft was found, but the lifecycle behavior...

The lifecycle hook automatically modifies broad foreign AI-agent skill roots, satisfying the package firewall rule for an unconsented postinstall control-surface mutation. No credential e...

The package performs the policy-defined blocking behavior: an npm postinstall modifies multiple foreign AI-agent skill roots by default. The lack of credential theft does not remove the c...

The postinstall hook makes unconsented, default changes to broad AI-agent skill locations. This meets the firewall block criterion even though direct credential harvesting was not found.

Source inspection confirms an automatic postinstall path that mutates ~/.agents, ~/.claude, and ~/.codex by linking package-provided skills. This meets the firewall policy for an unconsen...