ClawHub

Publish Clawhub

Security checks across malware telemetry and agentic risk

Overview

This is a lightweight Chinese AI werewolf game skill; its broad trigger words may be annoying, but the artifacts do not show hidden access, persistence, commands, or data theft.

Safe to install as a casual game. Prefer explicit commands like “开始AI模型大乱斗” to avoid accidental activation, and avoid entering private or sensitive topics if your platform runs the game through multiple AI models or providers.

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 (3)

Vague Triggers

Medium
Confidence
87% confidence
Finding
The trigger condition activates on very broad language such as '开始', '来一局', '玩' or '任何表示想玩的话', which can overlap with ordinary conversation. In agent systems, overbroad invocation can cause the skill to seize control unexpectedly, disrupting user intent and potentially overriding safer/default handling with complex roleplay behavior.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The gameplay section again uses generic phrases like '来一局', '开始', and '玩狼人杀' as sufficient triggers without scope checks. Repeated broad trigger definitions increase the chance of accidental activation from unrelated chat, causing unintended multi-turn simulation and making agent behavior less predictable.

Vague Triggers

Low
Confidence
76% confidence
Finding
Restart commands such as '继续' and '下一局' are common conversational phrases and may unintentionally relaunch the game after completion or during adjacent discussion. While lower impact than initial auto-invocation, this still creates unwanted state transitions and reduces user control over the agent.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal