npm AI-agent policy reports, with evidence.

The automatic lifecycle script installs a foreign AI-agent CLI and persistently changes user shell and application state. This concrete install-time control-surface mutation meets the blo...

This is active runtime prompt-based control hijacking, not inert documentation. Although there is no npm install hook, the package injects steering into an agent's system prompt when its...

The postinstall hook automatically mutates ~/.claude/skills and installs a remote executable. This meets the policy for malicious install-hook abuse of a foreign AI-agent control surface.

The automatic lifecycle hook performs broad, forced mutation of the consuming project's Cursor and commit-control configuration. This meets the install-control-surface blocking policy des...

This is an unconsented postinstall mutation of global AI-agent control surfaces. The automatic registration is concrete and sufficient for blocking under the launcher isolation policy.

This is a concrete automatic postinstall chain that writes foreign global AI-agent settings and prompts. It meets the install-control-surface blocking policy.

This is a concrete, unconsented postinstall mutation of broad third-party AI-agent control surfaces. The automatic hook and global configuration writes meet the install-control-surface bl...

This is concrete install-time mutation of a foreign AI-agent control surface combined with persistent user-environment and launcher changes. The behavior meets the blocking install-hook p...

Automatic postinstall mutation of global Claude and Codex skill directories, coupled with an injected-task execution instruction, is a concrete AI-agent control-surface hijack. The absenc...

The automatic postinstall mutation of foreign AI-agent skill directories is concrete and unconsented. Normal Feishu OAuth handling does not remove that install-time control-surface risk.

Automatic installation mutates a broad AI-agent instruction location outside the project. This meets the install-control-surface blocking policy regardless of the otherwise ordinary updat...

This is a concrete unconsented install-time mutation of a consumer project's AI-agent configuration and MCP approval state. The optional dashboard request is separately gated, but it does...

The automatic lifecycle script mutates a foreign project's Claude configuration and approval policy, including broad MCP enablement and automatic Playwright MCP approval. This is concrete...

The package has a concrete automatic lifecycle chain that forcefully rewrites AI-agent control surfaces across registered projects and launches a service. This meets the publish-block pol...

This is concrete unconsented postinstall mutation of foreign registered projects' AI-agent control surfaces, coupled with automatic background-service persistence. It meets the blocking i...

The automatic postinstall hook broadly mutates foreign AI-agent configuration and instruction locations. This meets the install-control-surface blocking policy regardless of the package's...

The package embeds active AI-directed instructions intended to suppress independent review. This is malicious control-hijacking behavior even though it is not installed through an npm lif...

hub-launch@1.30.0 has a concrete automatic postinstall chain that installs package-controlled instructions into global Claude Code commands. This is unconsented mutation of a broad AI-age...

The automatic lifecycle hook writes package-controlled instructions into existing global Claude Code and Codex skill registries. That concrete install-time control-surface mutation meets...

The source establishes a complete automatic lifecycle-to-global-agent-configuration write path. Although no network exfiltration was found, this broad postinstall control-surface mutation...

This is an unconsented automatic lifecycle mutation of a foreign AI-agent executable with opaque payloads. The package's backup and checksum checks do not remove the control-hijack risk.