ClawHub
Back to skill

Security audit

深夜食堂

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only Chinese roleplay/chat skill with no code, install scripts, data access, or persistence.

Install this if you want a Chinese 'midnight diner' conversation persona. Be aware it may activate during general emotional-support or late-night chat prompts, and it is not a substitute for professional mental-health, medical, legal, or emergency help.

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
94% confidence
Finding
The skill declares very broad triggers such as wanting to chat, seeking comfort, sharing stories, or late-night casual conversation. This can cause the roleplay skill to activate in many ordinary conversations, overriding user intent and potentially interfering with safer or more appropriate system behaviors for emotional-support contexts.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The activation section explicitly triggers on vague contextual cues like low mood, anxiety, sharing troubles, and late-night chatting. In practice, these signals overlap with common support-seeking conversations, creating a risk of unsolicited persona takeover and reducing clarity about whether the user wanted roleplay versus direct assistance.

Natural-Language Policy Violations

Medium
Confidence
87% confidence
Finding
The skill content is written to enforce a Chinese-language roleplay persona without indicating that the user can choose another language or decline the roleplay format. This can cause unwanted language switching, reduce usability, and make the assistant behave inconsistently with the user's expressed preferences.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.