registry / cursor-browser-cli / 1.1.1

cursor-browser-cli@1.1.1

CLI + MCP to drive Cursor IDE's built-in Browser Tab from Grok Build, Claude Code, Codex, or shell

AI Security Review

scanned 5d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Install-time code mutates several foreign AI-agent skill roots and can register a Grok MCP server. It also installs a startup-activated Cursor extension.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

npm postinstall

Impact

Unconsented expansion of AI-agent control surfaces with browser automation capability.

Mechanism

cross-agent skill deployment and MCP/extension registration

Policy narrative

Installing the package automatically runs setup, which writes its skill into multiple agent ecosystems, installs a Cursor extension, and conditionally registers a Grok MCP command. The installed skill exposes browser automation, page JavaScript evaluation, console/network inspection, and screenshots through the local extension bridge.

Rationale

This is a concrete unconsented postinstall mutation of foreign and broad AI-agent control surfaces. Localhost-only transport does not mitigate the install-time cross-agent registration.

Evidence

package.jsonscripts/setup.jsskill/SKILL.mdextension/package.jsonextension/extension.jsmcp/server.mjsextension/snapshot.js~/.grok/skills/cursor-browser/SKILL.md~/.claude/skills/cursor-browser/SKILL.md~/.agents/skills/cursor-browser/SKILL.md~/.cursor/skills/cursor-browser/SKILL.md

cursor-browser-cli@1.1.1

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code