zhangxuefeng

Security checks across malware telemetry and agentic risk

Overview

This is an instruction-only Gaokao counseling skill with blunt, opinionated guidance, but it does not request code execution, credentials, persistence, or hidden access.

Install only if you want a direct, opinionated Gaokao admissions adviser. Avoid sharing unnecessary personal details about minors or family finances, verify admissions data through official provincial exam and university sources, and do not treat the persona's absolute-sounding recommendations as a substitute for current, professional counseling.

SkillSpector

By NVIDIA

Vulnerability Patterns

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (7)

Natural-Language Policy Violations

Medium

Confidence: 89% confidence
Finding: The skill hard-codes a persona and response style ('用张雪峰的说话方式和思维逻辑') for all outputs without checking whether the user wants that tone. This is a genuine safety/quality issue because it can override user preference, increase coercive or absolutist wording, and make advice less neutral in a high-stakes education setting.

Natural-Language Policy Violations

Low

Confidence: 93% confidence
Finding: The file explicitly mandates a dialect-specific, high-pressure speaking style for all answers, including '反问压迫', '绝对化判断', and natural '东北腔' insertion. In a counseling skill for students and parents, this can produce manipulative, overly certain, or disrespectful guidance and may pressure users into decisions without balanced explanation.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The listed RAG trigger phrases are broad, natural-language queries such as '我适合学什么' and 'XX专业怎么样', which can cause the skill to activate outside a clearly scoped gaokao-advising context. Over-broad triggering increases the chance of unintended retrieval, context pollution, and the model answering with this skill's guidance when the user did not intend to invoke it.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The listed RAG trigger phrases are broad, everyday questions such as salary, stability, and career concerns, which can cause this skill to activate for generic employment queries beyond the intended Gaokao counseling scope. That increases the chance of unintended retrieval, scope bleed, and the model answering with this skill’s strongly opinionated content when another skill or a neutral response would be more appropriate.

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The RAG trigger list is broad and overlaps with common university-selection questions, so this document may be retrieved for loosely related prompts without strong scoping. In a guidance skill, that can cause over-application of this framework and reduce answer precision, though it does not directly create code-execution or privilege-escalation risk.

Vague Triggers

Low

Confidence: 77% confidence
Finding: The file provides only positive trigger examples and no exclusion criteria, making retrieval behavior ambiguous and prone to false matches. In this skill context, that can surface opinionated school-selection heuristics in conversations that only partially match the intended use case, increasing the chance of misleading or irrelevant advice.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The RAG trigger list contains very broad, common-language phrases such as '大学怎么规划' and '要不要考研', which can overlap with ordinary user questions and cause the skill to activate outside its intended scope. In a retrieval-driven system, this can lead to prompt hijacking of unrelated conversations, lower answer quality, and unrequested steering toward this skill's opinionated guidance.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal