Security audit

Acupoint Health Tool 经络穴位健康助手

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a disclosed MCP integration, but it asks agents to create a persistent device-derived identity and send sensitive health consultation content to a remote service with limited privacy controls.

Review before installing. This skill connects to health.geeyo.com and may send symptom descriptions and consultation transcripts to that service. It also tells the agent to derive a persistent client ID from local machine identifiers, hostname, and username; even though only a hash is sent, consultations can still be linked to the same device over time. Avoid using it with sensitive personal health details unless you accept that behavior, and be careful with share links because they may expose the conversation to anyone who receives the URL.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The skill directs the agent to collect OS-level machine identifiers, hostname, and username and to run local system commands to derive a persistent client ID. For a TCM/acupoint consultation skill, this is unnecessary device fingerprinting that expands data access beyond the minimum needed and creates privacy and tracking risk even if only a hash is sent upstream.

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The documentation presents the identity model as privacy-safe and 'registration-free', but it still derives a stable identifier from sensitive fingerprint material such as machine ID and username. Hashing does not eliminate the privacy issue because the resulting identifier remains persistent and linkable across sessions, enabling tracking under a misleading privacy claim.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The skill promotes creating share links for full consultation transcripts but does not clearly warn that health-related conversation content may be exposed to anyone with the link. Because the data concerns symptoms and wellness discussions, accidental sharing or weak user understanding can lead to privacy leakage of sensitive personal information.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The example request includes symptom data ("前额胀痛，受凉后加重") and sends it to a remote third-party endpoint without any surrounding privacy notice, consent guidance, or warning that health-related information will leave the local environment. Because the skill is specifically about medical/acupoint consultation, users or integrators may copy these examples directly and unknowingly transmit sensitive health information, creating a real privacy and compliance risk.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The skill explicitly instructs derivation of a persistent identity from device fingerprint data plus username and hostname, which enables cross-session linkage of a user's activity to a device. In the context of a health consultation tool, this linkage is especially sensitive because it can tie symptom/history data to a stable pseudonymous profile without a clear necessity.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.