Security audit

物理实验教学指导

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a teacher-facing physics lab coaching guide, but it needs review because its safety and scoring boundaries are not consistently scoped.

Review before installing for classroom use. Treat its rubrics as teacher-only references, not automatic grading instructions, and do not reuse the extracurricular electrical experiment examples unless you add strict safety limits, adult supervision, and no direct interaction with household mains circuits.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (5)

Intent-Code Divergence

Medium

Confidence: 90% confidence
Finding: The skill declares that it will not score lab reports, but later provides a concrete weighted scoring rubric. This inconsistency can cause downstream agents or users to rely on the skill for evaluative decisions it claims not to make, weakening policy boundaries and enabling unauthorized or over-automated student assessment.

Intent-Code Divergence

Medium

Confidence: 82% confidence
Finding: The skill forbids producing a 'standard lab report', yet it references templates and report samples detailed enough to be used as turnkey report outputs. In practice, this can let users obtain prebuilt reports or near-complete writeups, undermining the stated restriction and encouraging academic misuse or unauthorized completion of student work.

Vague Triggers

Medium

Confidence: 78% confidence
Finding: The activation phrases are broad and match common classroom conversation such as asking how to teach experiments or write reports. Overbroad routing can invoke the skill in unintended contexts, causing unnecessary data flow to connected skills or producing responses outside the teacher's actual intent.

Vague Triggers

Low

Confidence: 74% confidence
Finding: The trigger section provides positive examples but no exclusions or negative examples, so the orchestration layer lacks guidance on when not to activate the skill. This increases the chance of accidental activation, especially in a teacher-support environment where similar phrasing is common across planning, grading, and classroom coaching tasks.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The document recommends extracurricular experiments such as measuring household circuit power and making an electromagnet, but it provides no user-facing safety constraints, supervision requirements, or prohibitions on dangerous procedures. In an education skill aimed at teachers, this can propagate unsafe classroom or at-home activities involving electricity, overheating, short circuits, or mains exposure, especially if teachers reuse the examples directly with students.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.