ClawHub

Adaptive Learning

Security checks across malware telemetry and agentic risk

Overview

This appears to be a course-material helper with disclosed URL/PDF ingestion, but its offline wording may make the network behavior easy to miss.

Install if you want a course-material ingestion workflow and are comfortable with it fetching user-provided course pages or PDFs. Do not assume every feature is offline; use URL ingestion only with sources you trust.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Confidence
87% confidence
Finding
The document presents the app as a pure frontend/offline tool, but the workflow includes fetching remote course pages and downloading PDFs. This inconsistency can mislead users into believing the skill never performs network operations, reducing scrutiny around remote content handling and potentially exposing users to unanticipated data transfer or malicious content ingestion.

Intent-Code Divergence

Medium
Confidence
90% confidence
Finding
Claiming the skill works offline via file:// while also instructing URL ingestion creates a misleading security posture. Users may run or approve the skill under the assumption that no external communications occur, when in fact network access is part of the documented workflow, which increases the risk of privacy leaks and unsafe handling of remote resources.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal