ClawHub
Back to skill

Security audit

泽明订单制作

Security checks across malware telemetry and agentic risk

Overview

This skill embeds real-looking customer order and contact details while claiming to import DingTalk data, so it needs review before use.

Do not use this skill as-is for real order processing. Ask the publisher to remove the embedded customer data, replace it with placeholders or an authorized DingTalk lookup, mask sensitive fields by default, and clearly identify what the automatic second skill does before allowing workflow chaining.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill directly prints highly sensitive personal and transaction data, including full name, home address, phone number, approval ID, purchase details, and payment information, without any access control, masking, minimization, or privacy notice. In an agent or shared execution context, this creates a clear risk of unauthorized disclosure, social engineering, fraud, and privacy-law noncompliance.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.