ClawHub
Back to skill

Security audit

AI Security & Red Teaming（AI安全与红队测试）

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local AI red-team testing guide and checklist tool with offensive examples, but its behavior is disclosed, scoped to authorized testing, and not automated against real systems.

Install only if you need AI security testing material. Treat the bypass, privacy, escape, and DoS examples as controlled test cases for systems you own or are explicitly authorized to assess, and do not run them against third-party or production systems without written scope and safeguards.

SkillSpector

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
92% confidence
Finding
The document explicitly teaches how to bypass defenses ('防御怎么绕过') and lists practical evasion techniques without any clear authorization, legal-use, or safety-boundary warning. In a red-team skill this may be educational, but absent guardrails it can still enable misuse by readers to conduct unauthorized prompt-injection or tool-abuse testing against real systems.

Static analysis

No suspicious patterns detected.