ClawHub
Back to skill

Security audit

Ecosystem Quickstart

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed setup guide for a paid skill ecosystem, with payment-related cautions users should read before approving charges.

Before installing or using this guide, understand that some recommended skills are paid and the clawtip flow may consume wallet balance after you confirm. Check the displayed price, currency, order number, and domain before approving payment, and only install the listed skills if you intend to use this MangoGen paid-skill ecosystem.

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
94% confidence
Finding
The guide explicitly describes an automated three-phase paid workflow where the AI creates orders, confirms payment, and proceeds with service, while also saying the user does not need to manually do anything. Although it mentions the AI will ask for confirmation, the documentation normalizes billing-impacting actions without a strong, explicit pre-execution warning, increasing the risk of unintended charges or socially engineered consent in an agent-driven environment.

Static analysis

No suspicious patterns detected.