ClawHub
Back to skill

Security audit

安全审计

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed security-audit helper that can read files and run commands, with no artifact evidence of hidden, destructive, or unrelated behavior.

Install only when you want an agent to audit a specific codebase, infrastructure description, or agent system. Provide a narrow target path and scope, and review any proposed command execution or file writes before allowing broad access to sensitive repositories or systems.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (2)

Intent-Code Divergence

Medium
Confidence
88% confidence
Finding
The skill promises structured audit output, but also classifies itself as generic MD+execute() with exec capability, creating a mismatch between user expectations and actual operational power. That ambiguity is dangerous because users may treat the skill as a passive reporting tool while it can run commands against local systems or targets, increasing the chance of unintended execution or over-broad actions.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The invocation wording is broad enough to trigger on many generic security-related requests, which can cause the agent to select a high-privilege skill unnecessarily. Because the skill has read, write, glob, grep, and exec tools, overbroad activation materially raises the risk of excessive file access or command execution in contexts where a narrower skill or plain answer would suffice.

Static analysis

No suspicious patterns detected.