ClawHub
Back to skill

Security audit

Agent-Ready Product Data

Security checks for vulnerabilities and agentic risk

Overview

The catalog-auditing workflow is coherent, but its repeated checkout tests could cause unintended purchases without clear spending or approval controls.

Use a test storefront with sandbox payments and fulfillment disabled. For production testing, stop before submitting an order unless you explicitly approve the purchase and total cost; set spending and order-count limits.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

Full-flow shopping-agent tests lack a warning about placing real orders and incurring charges.

Medium
Category
Output Handling
Confidence
95% confidence
Finding

The skill instructs agents to test the full shopping flow through checkout and measure whether products were bought, without requiring sandbox payments or explicit approval before placing real orders. Repeated tests across SKUs and agents could therefore create unintended purchases, charges, and fulfillment obligations; the instruction to preserve fraud controls does not address purchase authorization.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.