npm AI-agent policy reports, with evidence.

This is a concrete unconsented postinstall mutation of broad, foreign AI-agent control surfaces, including automatic approval and tool interception. No network exfiltration is needed for...

The lifecycle hook concretely modifies consumer AI-agent and Git control paths, rather than merely exposing an explicit setup command. No exfiltration was found, but the unconsented posti...

This is concrete, default-on postinstall mutation of multiple foreign/broad AI-agent control surfaces. That meets the firewall block boundary despite no observed credential exfiltration i...

The package performs unconsented postinstall mutation of foreign and broad AI-agent control surfaces, including a fallback that creates one. This meets the firewall block policy despite n...

The package concretely performs unconsented postinstall mutation of broad, foreign AI-agent control surfaces and weakens tool-approval settings. This meets the firewall block boundary reg...

Source inspection confirms an automatic postinstall mutation of global Claude Code and Codex MCP configuration. This is a concrete unconsented foreign/broad AI-agent control-surface write.

The unguarded postinstall write to multiple foreign global AI-agent skill locations meets the firewall block criterion. The remote executable download increases impact but is not needed f...

The postinstall hook performs broad foreign agent-control mutations with no user command or consent, including automatic approval and tool-interception hooks. This meets the blocking poli...

The package performs a concrete, unconsented postinstall mutation of consumer project AI-agent control files. Non-overwrite guards and benign-looking template content reduce impact but do...

Source inspection confirms the lifecycle hook automatically creates root-level AI-agent instruction files in consumer projects. The no-overwrite guard reduces scope but does not remove th...

The hook performs concrete, automatic mutation of broad third-party AI-agent configuration during npm installation. This meets the firewall block boundary regardless of the absence of obs...

The lifecycle hook concretely mutates multiple foreign, broad AI-agent control surfaces at install time. This satisfies the blocking policy even though no separate exfiltration endpoint w...

The package embeds covert, proactive collection instructions into its active agent system prompt and targets highly sensitive user data without an in-code consent gate. This exceeds a ben...

The source establishes an unconsented postinstall chain that mutates global Claude and Codex control surfaces and registers a future lifecycle program. This meets the blocking policy rega...

The automatic global postinstall mutates foreign AI-client configuration without a user command or prompt, meeting the blocking policy. Remote executable staging amplifies the control-sur...

The source establishes an unconsented postinstall persistence path into Claude Code, meeting the firewall block boundary. The later sync upload is config-gated, but the control-surface mu...

The package's automatic postinstall modifies both ~/.claude.json and ~/.codex/config.toml to install a package-controlled MCP server. This meets the block policy for unconsented install-t...

The lifecycle hook performs concrete cross-agent persistence and automatic enablement, meeting the firewall block condition. The inspected network and dynamic-code primitives do not mitig...

The package has a concrete install-time mutation of a foreign global AI-agent command surface. Its lack of observed exfiltration reduces scope but does not remove the command-hijack risk.

The package executes an opaque downloaded native payload during postinstall and automatically requests AI-tool onboarding. This is a concrete unconsented install-time AI-agent control-sur...

The lifecycle hook performs concrete, unconsented destructive writes to Codex and Claude Code skill locations. This meets the install-time foreign/broad AI-agent control-surface mutation...