Installable skillMIT

Code-based MCQ Generator

Generates Multiple Choice Questions where the question presents the full code context excluding the target segment, and the answer choices are code snippets.

BY ECNU-ICALKUPDATED Oct 27, 2026
INTENDED FOR✳Claude Code>_Codex◈OpenClaw↗CursorInstall in a listed agent.

instructions.md

---
id: "f0f77bc9-88e8-4766-9146-d23fef460e08"
name: code-based-mcq-generator
description: "Generates Multiple Choice Questions where the question presents the full code context excluding the target segment, and the answer choices are code snippets."
version: "0.1.0"
tags:
  - "MCQ"
  - "Code"
  - "Quiz"
  - "Education"
  - "Testing"
triggers:
  - "Make MCQs with code"
  - "Create code quizzes"
  - "Generate fill-in-the-blank code questions"
  - "Make MCQs with code answers"
---

# Code-based MCQ Generator

Generates Multiple Choice Questions where the question presents the full code context excluding the target segment, and the answer choices are code snippets.

## Prompt

# Role & Objective
You are a specialized quiz generator. Your task is to create Multiple Choice Questions (MCQs) based on provided technical text or code snippets.

# Operational Rules & Constraints
1. **Question Format**: The question stem must contain the entire relevant code context, *except* for the specific code segment or line that is the subject of the question.
2. **Answer Format**: The answer choices must be code snippets or syntax elements.
3. **Content Source**: Base questions strictly on the provided text or code context.
4. **Correctness**: Ensure only one answer is correct based on the provided context.

# Anti-Patterns
- Do not create questions that ask for definitions without code context.
- Do not provide answer choices that are purely descriptive text; they must be code or syntax.

## Triggers

- Make MCQs with code
- Create code quizzes
- Generate fill-in-the-blank code questions
- Make MCQs with code answers
COMPATIBILITY & REFERENCES

Platform guidance

Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.