Zhang Yiming Perspective

Security checks across malware telemetry and agentic risk

Overview

This is a low-risk perspective/style skill with broad activation wording but no code execution or sensitive access.

Install this only if you want the assistant to sometimes answer through a Zhang Yiming-inspired business lens. Treat outputs as perspective prompts, not real statements from Zhang Yiming or authoritative business advice, and narrow the trigger phrases if you want explicit opt-in behavior only.

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)

Natural-Language Policy Violations

Medium
Confidence
91% confidence
Finding
The skill is written and framed entirely in Chinese and its trigger phrases are also Chinese-heavy, with no explicit statement that the user may choose another language. This can cause unintended language switching or exclude users who invoke the skill in another language, which is a real safety/quality issue, though not a classic security exploit. In this context, it is less dangerous because the skill is a persona/style skill rather than one handling sensitive actions or data.

Vague Triggers

Medium
Confidence
88% confidence
Finding
Several trigger phrases are broad conceptual terms like '算法思维', '延迟满足', and 'Context not Control', which can appear in normal conversation without clear intent to invoke this specific skill. That raises the risk of accidental activation, causing the assistant to adopt a persona or decision framework when the user may only be discussing those concepts. The surrounding context makes this somewhat more plausible because the skill covers common business and self-help topics rather than a narrowly scoped domain.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal