ClawHub
Back to skill

Security audit

Human Level Up

Security checks across malware telemetry and agentic risk

Overview

This is a coherent learning and quiz skill, with the main caution that optional deployment examples can send selected or uploaded content to external services.

Installers should be comfortable with a Chinese-language learning workflow. Do not use the bookmarklet, Docker image, or serverless examples with private, regulated, or proprietary documents unless you have inspected the external endpoint or image and understand where the content is processed.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The README instructs users to upload documents and includes examples that send selected content to a remote API, but it does not clearly warn that user data may leave the local environment or describe privacy implications. This can lead users to transmit sensitive documents or selections without informed consent, especially in a learning tool that encourages arbitrary content submission.

Natural-Language Policy Violations

Medium
Confidence
95% confidence
Finding
The skill hardcodes Chinese-language behavior across the entire prompt without offering a user choice or fallback. This can exclude users, cause misunderstanding of safety-critical content, and create accessibility and usability issues when the user's preferred language is different.

Vague Triggers

Medium
Confidence
94% confidence
Finding
The trigger list contains very generic phrases such as “提取”, “重点”, “精华”, and “来一道”, which are likely to appear in normal user conversations unrelated to this skill. That increases the chance of accidental invocation or routing collisions with other skills, causing unintended behavior and possible exposure of user-provided content to the wrong workflow.

Natural-Language Policy Violations

Medium
Confidence
85% confidence
Finding
The skill text is primarily written in Chinese and frames the interaction style and goals in that locale without any visible user language negotiation or opt-in. While not directly a code-execution risk, this can cause misleading activation, poor user comprehension, and unintended behavior if users do not understand the instructions or the system assumes a language context they did not request.

External Transmission

Medium
Category
Data Exfiltration
Content
javascript:(function(){
  const text = window.getSelection().toString();
  if(text.length > 100) {
    fetch('https://your-api/challenge', {
      method: 'POST',
      body: JSON.stringify({content: text})
    })
Confidence
96% confidence
Finding
The bookmarklet example captures user-selected page text and posts it to a remote endpoint, creating a clear external data transmission path. In context, users may select proprietary, personal, or otherwise sensitive content from arbitrary webpages and transmit it without strong notice, minimization, authentication, or origin-specific safeguards.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.