ClawHub
Back to skill

Security audit

GEO Prompt Architecture

Security checks across malware telemetry and agentic risk

Overview

This skill is a documentation-heavy GEO prompt-generation helper with no executable code, hidden data handling, persistence, or credential use found.

Reasonable to install for GEO prompt architecture work. Provide only the client briefs, websites, screenshots, exports, or monitoring results you intend the skill to use, and be aware the README should clean up local filesystem links and some reference prompts may need an explicit language instruction if you do not want Chinese output structure.

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 (1)

Intent-Code Divergence

Low
Confidence
96% confidence
Finding
The README exposes absolute local filesystem paths under /Users/timlin/... while presenting them as example links. This leaks developer environment details and can disclose usernames, workstation layout, and repository locations that aid social engineering or targeted follow-on attacks. In a skill repository, this is not directly exploitable on its own, but it is still an avoidable information disclosure.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.