ClawHub
Back to skill

Security audit

Haoyunge

Security checks across malware telemetry and agentic risk

Overview

This is a text-only finance persona skill, so its main risk is overconfident trading-style commentary rather than hidden code or system access.

Install only if you want simulated Haoyunge-style trading commentary. Treat responses as educational or roleplay commentary, not as official statements, personalized financial advice, or instructions to trade; ask the assistant to exit the role or provide neutral risk analysis before making investment decisions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The manifest presents the skill as a perspective or expression framework, but the trigger phrases and body content explicitly steer the model toward concrete stock selection, timing, and suitability judgments. This mismatch can mislead users and downstream safety systems into treating advisory behavior as harmless roleplay, increasing the chance of unvetted financial guidance being delivered.

Description-Behavior Mismatch

Medium
Confidence
89% confidence
Finding
The skill requires first-person impersonation of a real investor and instructs the model to avoid meta-analysis, which can cause users to perceive the output as authoritative personal advice rather than stylized commentary. In a financial context, this increases deception and trust risks, especially when paired with strong, confident judgments and minimized disclaimers.

Intent-Code Divergence

High
Confidence
97% confidence
Finding
The disclaimer claims the skill is only for mindset training and not investment advice, but the skill repeatedly invites direct judgments like whether a stock is worth doing, whether it is a leader, and whether now is a good time. A contradictory disclaimer can create false reassurance while the system still produces actionable financial guidance, making harmful reliance more likely.

Vague Triggers

Medium
Confidence
83% confidence
Finding
The trigger list contains broad, ordinary phrases such as asking whether now is a good time or whether the bull market is far away, which can cause accidental activation in general market discussions. In this skill, accidental activation matters because it can unexpectedly switch the assistant into a high-confidence financial persona that provides actionable trading-style guidance.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.