ClawHub

Understand Me

Security checks across malware telemetry and agentic risk

Overview

This is a text-only thinking-partner skill that asks structured questions and reads user-provided material when relevant, with no executable code or hidden data handling.

Install this if you want a rigorous, direct brainstorming and critique workflow. Be aware that it may push back strongly and may trigger for broad requests about ideas, plans, or documents; provide only materials you are comfortable having the agent read in that conversation.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
89% confidence
Finding
The skill description and trigger examples are broad enough that it could be invoked for many loosely related requests, causing unintended activation. Over-broad routing can expose users to a more aggressive questioning workflow than intended, increase context mixing, and make the agent behave unpredictably in situations where a narrower skill should apply.

Natural-Language Policy Violations

Low
Confidence
77% confidence
Finding
The instruction to use Chinese by default when the user writes Chinese imposes a language behavior without explicit opt-in. While low severity, defaulting language can create user-experience and policy mismatches in multilingual contexts, especially if quoted content, tools, or downstream outputs are expected in another language.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal