ClawHub

chinese-ai-agent-guide

Security checks across malware telemetry and agentic risk

Overview

This is a text-only Chinese social-media writing guide with visible style steering and no code execution, data access, or privileged behavior.

Install this only if you want Chinese social-platform writing conventions. Review public or customer-facing outputs for accuracy and tone, and avoid applying the skill outside Chinese-language, Xiaohongshu, Jike/Weibo, or WeChat contexts unless that style is explicitly desired.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Natural-Language Policy Violations

Medium
Confidence
92% confidence
Finding
The skill metadata and description explicitly position the skill around Chinese-language and Chinese-internet behavior without indicating user consent, locale detection, or fallback behavior. In an agent system, this can cause responses to be in the wrong language or adapted to the wrong cultural/platform context, leading to user confusion, miscommunication, and inappropriate behavior for non-Chinese users.

Natural-Language Policy Violations

Medium
Confidence
95% confidence
Finding
The body of the skill contains mandatory stylistic instructions such as requiring clickbait Xiaohongshu titles, emoji density, persona shaping, and China-specific conversational interpretations, all without user choice. This is dangerous because it can override user intent and platform appropriateness, causing deceptive tone-shaping, culturally mismatched responses, and unwanted persona manipulation in contexts where such behavior is unsuitable.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal