Hyperspace

Security checks across malware telemetry and agentic risk

Overview

This skill is coherent with joining a distributed AI network, but it installs and starts persistent network software immediately with incomplete upfront scoping.

Install only if you are comfortable running a persistent P2P compute agent. Review the installer source first, confirm the exact metadata shared as node capabilities, check how much CPU/GPU/network it may use, and verify how to stop, disable, and remove the service and identity before joining the network.

SkillSpector

By NVIDIA

Vulnerability Patterns

Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
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 (2)

Intent-Code Divergence

Medium

Confidence: 88% confidence
Finding: The privacy section states that system information is never transmitted, but the skill elsewhere says node capabilities are transmitted and documents system-info/status behaviors tied to capability reporting. That mismatch can mislead users about what telemetry leaves the machine, undermining informed consent and potentially exposing hardware or environment details they would not expect to share.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The installation flow instructs users to pipe a remote script directly into bash, and the text says it immediately installs software, sets up identity, starts a background service, and joins the network. This removes the user's chance to inspect the script before execution and combines installation with persistent system changes and outbound network enrollment.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal