ClawHub
Back to skill

Security audit

Qubitclient Control

Security checks across malware telemetry and agentic risk

Overview

This is a documentation-only skill for quantum measurement control, with no hidden execution or persistence found.

Install only in a lab environment where the MCP server is trusted and scoped to approved measurement tasks. Review server-side permissions and hardware safety controls before allowing agents to run parameter sweeps or experiments.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Description-Behavior Mismatch

Low
Confidence
86% confidence
Finding
Documenting direct MCPClient.call task invocation by arbitrary string name broadens the effective interface beyond the enumerated task wrapper and can enable access to any exposed MCP server operation, not just the intended fixed measurement tasks. In a hardware-control skill with real-time feedback and experimental execution, this increases the risk of invoking unintended or unsafe actions if the backing MCP server exposes more capabilities than the documentation summary suggests.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.