Security audit

Skill Customs

Security checks across malware telemetry and agentic risk

Overview

This customs-data lead generation skill is coherent, but it needs review because it can enrich leads and trigger outreach through business channels with under-specified controls and privacy boundaries.

Install only if you are comfortable sending customs-data queries and business-intelligence inputs to Yunlv/TradeGPT APIs and storing lead and outreach records locally. Keep outreach in manual review mode: approve every recipient, message, channel, and follow-up schedule yourself, and verify GDPR, CAN-SPAM, platform rules, and the accuracy of any competitor or import-volume claims before contacting prospects.

SkillSpector

By NVIDIA

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 (5)

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The skill makes contradictory privacy claims: it says query conditions and intermediary enterprise data are processed only locally, while the documented workflow and metadata explicitly require sending queries to Yunlv external APIs and customs-data services. This can mislead users and operators about where potentially sensitive business-intelligence inputs are transmitted, creating privacy, compliance, and trust risks.

Description-Behavior Mismatch

Medium

Confidence: 86% confidence
Finding: The file contains generic outbound sales and prospecting email templates, including a template explicitly framed around using customs-import data to target a company's existing supplier relationship. That expands the skill from trade-intelligence retrieval into actionable customer-mining and outreach enablement, which increases privacy, compliance, and abuse risk because the skill can operationalize potentially sensitive trade data into targeted solicitation.

Vague Triggers

Medium

Confidence: 82% confidence
Finding: The trigger list includes broad phrases such as '竞品分析', '采购记录', 'customs data', and 'trade intelligence', which can match user requests outside the narrow intended use case. Over-broad activation can cause the agent to invoke this skill in the wrong context, unnecessarily routing user queries and potentially sensitive commercial information to external APIs or lead-generation workflows.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger phrase "trade intelligence" is broad and can match ordinary business conversations that are not clearly requesting this specific skill. That increases the chance of unintended invocation, which can route user queries to an external customs-data service and expose business-sensitive prompts unnecessarily.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The description defines activation in very broad terms, such as any user needing import/export data or competitor customer mining, without clear boundaries or user-confirmation requirements. In a skill that calls a third-party API for trade intelligence, this can cause over-activation and unnecessary sharing of commercially sensitive user requests.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.