Ma Huateng Perspective V2

• Outcome: Review
• Latest audit: May 29, 2026 at 9:32 PM UTC
• Version:
  1.0.0

Overview

This is an instruction-only persona skill with no code access, but it strongly encourages first-person impersonation of Pony Ma and may activate without a clear request for that persona.

Install only if you intentionally want a strong simulated public-figure persona. Treat all outputs as AI-generated analysis, not statements from Ma Huateng or Tencent, and prefer explicit activation plus a visible disclaimer when using it for business, strategy, or investment discussions.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 96% confidence
• Finding: The skill is presented as a perspective/advisory framework, but the instructions require first-person identity impersonation of a real public figure. This can mislead users into believing the assistant is speaking as Pony Ma rather than providing a clearly labeled simulation, increasing risks of deception, authority laundering, and user confusion in strategic or investment discussions.

Intent-Code Divergence

Medium

• Confidence: 97% confidence
• Finding: The skill acknowledges that it is a synthesis from public information, yet simultaneously instructs the model to present itself directly as Pony Ma. That contradiction makes the output inherently misleading because it masks conjecture and reconstruction as authentic first-person statements from a real individual.

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The trigger conditions are broad enough to activate on generic product or strategy questions that do not clearly request this persona. This can cause unwanted takeover of normal conversations, unexpectedly shifting the assistant into a roleplaying mode and increasing the chance of deceptive or irrelevant outputs.

Natural-Language Policy Violations

Medium

• Confidence: 83% confidence
• Finding: The skill forces responses in the persona's Chinese voice without checking user language preference or obtaining opt-in. While not directly enabling code execution or data theft, it can degrade usability, override user intent, and make deceptive persona framing harder for some users to detect or challenge.

VirusTotal

66/66 vendors flagged this skill as clean.