ClawHub

罗永浩视角（升级版）

Security checks across malware telemetry and agentic risk

Overview

This is a text-only Luo Yonghao-style persona skill with disclosed roleplay behavior and no executable code, data access, persistence, or install-time actions.

Install this only if you want a roleplay-style advisor. Treat outputs as AI-generated interpretation based on public material, not as statements from or endorsed by Luo Yonghao. For shared or user-facing deployments, consider stricter explicit activation and a repeated visible disclaimer.

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 (2)

Vague Triggers

Medium
Confidence
92% confidence
Finding
The trigger phrases are broad enough that ordinary conversation about "how Luo Yonghao would think" or "switch to Lao Luo" could activate the skill without a clear, explicit user request for roleplay. Unintended activation can override the assistant’s normal response mode, increasing the chance of confusing users, bypassing expected tone/safety defaults, or causing the model to answer in a forceful persona when the user only wanted analysis.

Natural-Language Policy Violations

Medium
Confidence
95% confidence
Finding
The skill mandates direct first-person impersonation and explicitly prohibits stepping out of character except on user command, which can suppress normal transparency and make generated claims sound like they come from the real person. This is more dangerous in context because the persona is intentionally highly assertive and discourages meta-analysis, increasing the risk of misleading authority, reduced consent, and difficulty correcting unsafe or inaccurate outputs once activated.

VirusTotal

60/60 vendors flagged this skill as clean.

View on VirusTotal