ClawHub

Chrome Debug Launcher

Security checks across malware telemetry and agentic risk

Overview

The skill is meant to launch Chrome for debugging, but it can close all existing Chrome windows without confirmation and open a debug-enabled browser.

Install only if you are comfortable with the agent controlling local Chrome. Before use, confirm there is no unsaved browser work, prefer a version that uses a separate Chrome profile without killing existing sessions, and avoid leaving the debug browser running longer than needed.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The skill claims to launch two Chrome instances, including a normal one 'alongside' a debug instance, but its first action forcibly terminates all existing Chrome processes system-wide. This can destroy active user sessions, interrupt unrelated work, and cause data loss risk from abruptly closed tabs or forms. In context, the discrepancy increases danger because a user asking to open browsers would not reasonably expect all running Chrome instances to be killed first.

Intent-Code Divergence

Low
Confidence
90% confidence
Finding
The documentation states the skill opens two independent Chrome instances, but the procedure requires killing all running Chrome beforehand, which contradicts the advertised 'alongside' behavior. This mismatch can mislead users and downstream agents into executing a destructive action they did not consent to, increasing the chance of unsafe automation outcomes.

Vague Triggers

Medium
Confidence
83% confidence
Finding
The activation phrase includes 'or any similar request,' which is overly broad and can cause the skill to trigger on ambiguous or loosely related user prompts. In this skill's context, broad matching is more dangerous because activation leads to disruptive local system actions, including launching a remote-debug-enabled browser and terminating Chrome processes.

Missing User Warnings

High
Confidence
98% confidence
Finding
The skill forcibly kills all Chrome processes without warning, confirmation, or scope limitation. This is dangerous because it can abruptly terminate active browsing sessions, lose unsaved work, and log the user out of workflows; combined with the subsequent launch of a debug-enabled browser on port 9222, it creates a high-risk automation path from a simple natural-language trigger.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal