ClawHub
Back to skill

Security audit

需求解构

Security checks for vulnerabilities and agentic risk

Overview

This skill is a requirement-analysis helper with limited, disclosed read and URL-fetch capabilities and no executable payload.

Use public or approved requirement URLs, and mask production identifiers, customer data, financial records, screenshots, ID numbers, and phone numbers before analysis. Treat URL fetching as a network action that may contact the linked site even though the skill does not show hidden exfiltration behavior.

SkillSpector

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

High
Confidence
83% confidence
Finding
The skill states that data is not externalized and only written under workspace/, yet it explicitly allows WebFetch and accepts URL retrieval as input. That mismatch can mislead users into supplying sensitive requirement documents under a false privacy assumption, and can cause the agent to transmit request metadata or fetched content to external systems contrary to the stated trust boundary.

Static analysis

No suspicious patterns detected.