ClawHub
Back to skill

Security audit

测试数据工程

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed test-data engineering guide with sensitive database and masking examples, but its behavior is purpose-aligned and framed for controlled test use.

Use this skill only with approved test environments and documented permission before touching production-derived data. Prefer synthetic data, validate masking before export, and manually review any SQL DELETE, API mutation, Docker, or Bash command before running it.

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
95% confidence
Finding
The skill explicitly states it may use real production data for testing after masking, but the top-level description lacks strong authorization, minimization, and legal-governance guardrails. In a test-data engineering context this is dangerous because it can normalize copying production data into non-production environments, where masking may be incomplete or reversible and privacy obligations may be violated.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.