Security audit

XRRobot ROS2

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed local ROS 2 robot-control skill whose motion, camera, navigation, and grasping abilities match its purpose, with real-world safety precautions still required.

Install only on the intended XRRobot host. Treat every motion, grasp, navigation, follow, lane-keeping, autopilot, and orchestration action as real physical robot control: keep the area clear, supervise the robot, keep a stop path available, and expose web interfaces only on trusted networks. Avoid camera capture or YOLO data collection in sensitive spaces unless everyone affected has consented.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (7)

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The document provides concrete motion commands that can immediately move a physical robot, but it does not prominently warn operators to verify a clear area, maintain an emergency stop path, and ensure the robot is safe to actuate before issuing commands. In a robotics skill, omission of explicit physical-safety guidance increases the chance of accidental collision, pinching, or property damage when users copy commands directly.

Missing User Warnings

Low

Confidence: 95% confidence
Finding: The skill explicitly captures the current camera frame and returns a JPEG for agent analysis, but the description does not include a clear user-facing privacy warning or consent notice. This can lead to inadvertent collection and transmission of sensitive scene imagery, especially in real-world environments containing people, screens, documents, or private spaces.

Missing User Warnings

High

Confidence: 95% confidence
Finding: The visual grasp workflow launches base and grasping components without an explicit warning at that step that the robot may immediately move its manipulator or chassis. Although the document has general safety prerequisites earlier, users following only the task snippet could initiate hazardous physical motion without a fresh, localized warning or confirmation.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The task orchestration workflow combines navigation, grasping, and transport, which can cause autonomous movement and manipulation, but the section lacks a direct warning and confirmation gate before launch. Because this is a multi-stage autonomous workflow, the absence of a local safety warning increases the chance of unintended operation in an unsafe environment.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The document provides direct commands to kill ROS processes, launch robot subsystems, and execute grasp-related workflows without any operator safety gating, environment checks, or warnings about physical motion hazards. In a robotics control skill, these instructions can lead to unintended robot movement or manipulator actuation if followed blindly, especially in shared or unprepared spaces.

Missing User Warnings

Low

Confidence: 88% confidence
Finding: The guide instructs users to open a local orchestrator web interface that can control multi-step robot actions, but it does not warn that this interface may expose a powerful control surface if reachable from untrusted networks. In the context of a robot orchestration service, lack of network-scope and access-control guidance increases the risk of unauthorized access or unsafe remote triggering.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: This CLI directly exposes motion-capable commands such as move, turn, stop, and navigate without any user-facing confirmation, arming check, environment validation, or emergency-safety interlock in the wrapper itself. In a robotics control skill, that makes accidental invocation, misuse by an upstream agent, or unsafe operation in an occupied environment materially dangerous because it can cause unintended physical motion.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.