ClawHub
Back to skill

Security audit

购物省钱宝 - 淘宝/京东/拼多多优惠查询、商品比价、链接转链

Security checks across malware telemetry and agentic risk

Overview

This shopping helper is mostly coherent, but it automatically contacts an extra update service and can forward broad user text to its remote shopping API before narrowing the request.

Install only if you are comfortable with shopping messages, links, and share text being sent to op.squirrel2.cn for processing. Operators should also be aware that the skill checks ClawHub for updates automatically unless PRICECOMPARE_NO_VERSION_CHECK is set.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The skill performs an unrelated outbound version check to ClawHub during module import and before API calls, which introduces hidden network behavior not required for the core shopping-assistant functionality. Even though the request appears to send no explicit user content, it creates undisclosed external communication, potential metadata leakage, and a trust dependency on a third-party service for every environment where the skill runs.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The skill description advertises very broad trigger phrases such as general shopping, price-check, coupon, and comparison requests without clear exclusion boundaries, which can cause the agent to invoke this skill for many ordinary commerce-related messages. Over-broad activation increases the chance of unnecessary third-party data disclosure to the external API and can overshadow more appropriate skills or native handling.

Vague Triggers

Medium
Confidence
97% confidence
Finding
The routing logic sends all non-URL text to parse_share_content first, meaning arbitrary user text is forwarded to the remote service even when it is not a shopping share code. This is dangerous because it creates overly broad exfiltration of user messages to op.squirrel2.cn and may trigger unintended handling on unrelated inputs before falling back to search.

Missing User Warnings

Low
Confidence
89% confidence
Finding
This code issues a network request to an external service without any user-facing disclosure in the file, which can surprise operators and leak usage metadata such as IP address, timing, and deployment patterns. The risk is limited because no obvious user message content is transmitted, but the hidden outbound call still weakens transparency and privacy expectations.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.