my skill

Security checks across malware telemetry and agentic risk

Overview

This skill analyzes private chats as advertised, but it also pushes sensitive profiling into manipulation, forced negative labels, and style imitation without adequate user control.

Install only if you are comfortable with a Review-level skill that profiles people from chat records. Use it only with consent and lawful authority over the chats, avoid uploading private screenshots unless you trust the helper tools, treat the output as speculative, and do not use the trigger, dark-trait, or imitation sections for harassment, manipulation, impersonation, or decisions affecting someone else.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 95% confidence
Finding: The template explicitly asks for content such as how to gain the target person's attention by exploiting their '雷点' and speculative boundary-crossing scenarios ('一直黏在一起，他会做什么'). That goes beyond neutral personality analysis and facilitates manipulation, harassment, or coercive social engineering based on inferred vulnerabilities from private chats.

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: The 'Dark Trait Scanner' section requires the model to assume the subject has potential for '黑化' and to list at least three evil tendencies even when unsupported. This drives defamatory, ungrounded psychological labeling and encourages harmful inferences presented as analysis, increasing risk of reputational harm, bias amplification, and abusive downstream use.

Context-Inappropriate Capability

Medium

Confidence: 94% confidence
Finding: The '风格模仿指南', catchphrase list, communication tactics, and '模拟回复' move the skill from analysis into operational impersonation and message generation in another person's style. In the context of a personality-analysis skill built from chat logs, this materially increases the risk of deceptive imitation, fraud, relationship manipulation, and privacy-invasive profiling.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The skill advertises broad trigger phrases like '分析一下这个人' and '帮我分析一下TA', which can match many ordinary conversations and cause the agent to invoke a highly sensitive psychological profiling workflow unexpectedly. In this context, accidental activation is more dangerous because the skill performs deep personality and psychological inference on chat records, increasing privacy and consent risks.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill instructs the agent to read local files, upload local images to another tool to obtain a URL, and generate a PDF output, but it does not require explicit user confirmation or warn that files may be accessed, uploaded, and new output artifacts created. This is risky because the workflow handles potentially sensitive chat logs and may transmit or transform private data without sufficiently informed consent.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal