ClawHub
Back to skill

Security audit

mayubench

Security checks across malware telemetry and agentic risk

Overview

MayuBench is a documentation-only AI behavior benchmark, not a tool that installs code or accesses user data.

Install this as a benchmark/reference skill. Use automatic Arena-style evaluation only in deliberate test sessions, and avoid feeding real private user data to external judge models unless you have reviewed that separate evaluation setup.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
88% confidence
Finding
The line claiming that, once loaded, behavior-style questions will automatically trigger MayuBench evaluation is overly broad and underspecified. In an agent/skill system, ambiguous auto-invocation criteria can cause unintended activation, surprise routing of user inputs, or evaluation logic being applied outside the user's clear intent, which can disrupt normal operation or create prompt-routing abuse opportunities.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.