ClawHub
Back to skill

Security audit

exam-question-generator

Security checks across malware telemetry and agentic risk

Overview

This is a coherent exam-practice generator that reads study materials and writes local question files, with some usability cautions around broad prompts and overwriting generated outputs.

Install this if you want local practice exams generated from your study materials. Review generated files before sharing them, avoid using sensitive source notes unless you are comfortable with them appearing in outputs, and remember that “重新出题” can replace earlier generated question files.

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 (3)

Vague Triggers

Medium
Confidence
87% confidence
Finding
Using a very broad trigger like “继续” can cause the skill to run when the user did not specifically intend to invoke it. In this skill, execution performs file reads/writes and downstream skill calls, so accidental activation can overwrite artifacts or generate content without clear user consent.

Vague Triggers

Medium
Confidence
91% confidence
Finding
Accepting generic replies like “要/好/继续/是” as action triggers makes unintended chaining into the next skill highly likely during normal conversation. Because this response path can automatically trigger exam-rescue-html or other follow-on actions, the ambiguity increases the risk of unauthorized workflow progression and unwanted file generation.

Missing User Warnings

High
Confidence
93% confidence
Finding
The skill explicitly states that “重新出题” overwrites old files and also forbids asking for confirmation, creating a clear destructive-action path without user verification. In a conversational system where short commands may be issued casually or triggered unintentionally, this can lead to loss of prior generated content and integrity issues in workspace artifacts.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.