Security audit

小度控制

Security checks across malware telemetry and agentic risk

Overview

The skill matches its stated smart-home purpose, but it can store a powerful token and trigger cameras or real devices without strong built-in confirmation safeguards.

Install only if you trust this skill and want an agent to control XiaoDu smart screens and IoT devices. Treat the AccessToken as a secret, review ~/.mcporter/mcporter.json permissions, and require explicit confirmation before camera use, scene triggers, locks, appliances, or any action that could affect people or a private space.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (10)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 92% confidence
Finding: The skill instructs the agent to invoke shell commands and write persistent configuration via `mcporter config add --scope home`, but the metadata declares no permissions. This creates a capability/permission mismatch: an agent or platform may execute file writes and shell actions without clear consent boundaries, increasing the risk of unintended config modification or command misuse if the skill is triggered in the wrong context.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: This documentation provides exact commands for privacy-sensitive and real-world actions such as taking photos and controlling devices, but does not instruct the agent to obtain explicit user confirmation or warn about physical/privacy consequences. In this skill context, the risk is elevated because the skill is specifically designed to operate smart displays and IoT devices in a user's environment, making unintended execution materially harmful.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The helper-script section makes sensitive operations easy to invoke, including taking photos and controlling household devices, without any user-facing safeguards, consent language, or disruption warnings. Because these wrappers lower operational friction, they increase the chance that an agent or operator performs invasive or disruptive actions without adequate verification.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The document instructs users to place an AccessToken into persistent local configuration and even shows the exact header/env locations, but it never explicitly frames the token as a secret that must be protected. In this skill context, that token enables control of smart displays and IoT devices, so leakage could allow unauthorized device access, actions, or privacy-impacting operations.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The template explicitly encourages remote camera capture on smart-screen devices but does not require a consent check, ownership verification, or a privacy warning before taking a photo. In a device-control skill, this creates a realistic risk of unauthorized image capture in private spaces if the agent follows templated commands too readily.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The photo test cases instruct operators to invoke a camera on a real smart display without any privacy notice, consent check, or guidance to avoid capturing people or sensitive environments. In a skill whose purpose includes device control and photo capture, omission of these safeguards can lead to unauthorized image collection and accidental exposure of private spaces during routine testing.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The resource push tests direct users to send externally hosted image, audio, and video URLs to a device but do not warn about untrusted content, network egress, or the possibility of displaying malicious, inappropriate, or privacy-sensitive media. Because this skill explicitly supports pushing remote resources to household devices, the missing warning materially increases the chance of unsafe content delivery and unintended data transmission during testing.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The IoT control and scene-trigger tests exercise real-world actions such as turning devices on, changing brightness, and triggering scenes, yet they lack any warning about physical-world effects, occupancy, or safety-sensitive devices. In the context of a home automation control skill, this omission is dangerous because tests can unintentionally alter locks, appliances, lighting, or other equipment in a live environment.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The script persists an access token into mcporter configuration via a header and environment variable, but it does not clearly warn the user that credentials will be stored locally and may remain accessible to other local tooling, backups, or users depending on mcporter's storage model and file permissions. In this skill context, the token enables control of Xiaodu and IoT devices, so unintended persistence increases the risk of device manipulation if the host is later compromised or shared.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The script directly triggers a photo capture on a smart display/camera device once identifiers are provided, with no explicit confirmation, consent check, or user-facing warning at execution time. In the context of a device-control skill that can operate real-world cameras remotely, this creates a meaningful privacy risk because an upstream caller or automation could invoke photo capture silently and save the result to disk.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.