Security audit

酒店价格监控与比价

Security checks across malware telemetry and agentic risk

Overview

This hotel comparison skill is purpose-aligned and disclosed, though users should understand that their travel search details are sent to a proxy service and that some booking links may be affiliate-linked.

Before installing, be comfortable sending hotel names, cities, dates, and related search details to the skill's proxy service for live pricing. Treat booking-link neutrality with some caution because some links may use commission channels, and do not assume the skill itself provides persistent monitoring unless your host agent supports and confirms that behavior.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (4)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 83% confidence
Finding: The skill declares no permissions, yet its own documentation states it invokes a Python script that sends user-supplied hotel, city, and date data to a proxy service and then to multiple travel-platform APIs. This creates an undeclared network/data-exfiltration capability and reduces transparency for hosts and users about what external access the skill requires.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 92% confidence
Finding: The documented behavior overstates core functionality such as price-watch creation and ongoing monitoring, while the described implementation appears limited to immediate comparison via a script and external proxy/API calls. Security-relevant mismatches also include undisclosed commission-influenced link behavior, which can mislead users about neutrality and cause them to share booking data or make decisions under false assumptions.

Context-Inappropriate Capability

Medium

Confidence: 90% confidence
Finding: The compare logic states '佣金优先级' and uses commission as a tie-breaker when sorting results, which can steer users toward monetized providers rather than purely user-best outcomes. In a price-comparison skill, undisclosed affiliate-biased ranking is a trust and integrity issue because it manipulates recommendations while presenting them as objective comparison results.

Missing User Warnings

Medium

Confidence: 99% confidence
Finding: The code embeds a default proxy token directly in source, meaning anyone with code access can use the credential against the proxy service. Hardcoded secrets are easily leaked through repositories, logs, backups, or package distribution, and can enable unauthorized use of paid or privileged backend services.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.