prompt-zs

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed AI image/video prompt-writing helper with no evidence of hidden access, persistence, credential use, or destructive behavior.

Install this only if you want a Chinese-oriented AI image/video prompt assistant that may produce structured JSON with both Chinese and English fields. Be aware that broad terms like prompt optimization, image generation, or video generation may route requests into this skill even when you only meant to discuss those topics.

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 (5)

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The statement encouraging use of broad trigger phrases such as prompt optimization, AI drawing, image generation, and video generation can cause the skill to activate for a wide range of ordinary user requests. In an agent environment, overbroad invocation increases the chance of misrouting tasks, overriding more appropriate tools, or unnecessarily exposing user input to this skill.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The changelog explicitly says more trigger keywords were added to improve activation rate, which signals optimization for breadth rather than precision. In skill-routing systems, this can lead to unintended invocation on unrelated requests, reducing control over tool selection and creating a prompt-scope abuse risk.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The skill metadata says it must be used whenever users mention very common phrases like '生成提示词', 'AI绘画', '图像生成', or '视频生成'. These broad mandatory triggers can cause unintended invocation and override more appropriate skills or general assistant behavior, creating routing confusion and increasing the chance of policy-inappropriate or low-context outputs.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The fallback rule routes all remaining 'drawing/image needs' to mode 1 without defining boundaries or exceptions. This can capture unrelated requests, suppress clarification, and produce structured prompt output even when the user wanted analysis, safety guidance, editing help, or non-generation assistance.

Natural-Language Policy Violations

Medium

Confidence: 88% confidence
Finding: The required JSON schema forces an English field alongside Chinese output regardless of the user's stated language preference. This can cause the assistant to reveal extra content the user did not request, reduce controllability, and conflict with higher-priority instructions to answer in a single requested language or minimal format.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal