ClawHub
Back to skill

Security audit

Vscode Node Tool Free

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed VSCode/Cursor remote-IDE helper, but users should treat its file write/delete capability as real and keep it scoped to trusted workspaces.

Install only if you intend to let an agent operate a connected VSCode/Cursor workspace. Use read-only commands first, confirm the target node and workspace before writes or deletes, and verify that your gateway enforces path restrictions and command allowlists.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Intent-Code Divergence

Low
Confidence
87% confidence
Finding
The skill claims that absolute paths and ../ traversal will be blocked, but the file is only documentation and contains no enforcement mechanism. This can create a false sense of safety, leading agents or users to trust dangerous file operations that may not actually be constrained by the implementation behind the skill.

Vague Triggers

Medium
Confidence
82% confidence
Finding
The trigger condition '需要代码生成、编程辅助、调试测试、开发部署时使用' is broad enough to match many normal development requests, increasing the chance that the skill is invoked when its write/exec capabilities are unnecessary. Overbroad auto-selection raises the risk of unintended file modification or command execution in contexts where a read-only or safer skill would suffice.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill advertises write, edit, delete, and command-execution-adjacent capabilities without a prominent warning that it can change files or invoke external commands. In an agent ecosystem, missing consent and safety language materially increases the chance of destructive actions, especially because the metadata exposes read/write/exec/glob/grep tools.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.