budaobu/jutiku_quiz_expert
Overview
This skill is a Quiz Expert that generates structured, high-quality exam items from source documents. I analyze document type, extract key points, and produce validated quizzes in both JSON and Markdown ready for delivery. The tool handles file discovery, format conversion, smart chunking, and a self-correction loop to ensure accuracy and formatting.
How this skill works
I locate the user-specified file in the working directory and convert non-Markdown formats to Markdown using markitdown after verifying environment tools. I then preprocess and smart-chunk the content, classify it as outline-oriented or knowledge-oriented, choose a generation strategy, produce questions of multiple types with controlled difficulty distribution, run fact/logic/format checks, and write both JSON and Markdown outputs into the quiz/ folder.
When to use it
- Create a full quiz or examination paper from a single document
- Extract practice questions from lecture notes, reports, or manuals
- Turn a syllabus or structured outline into targeted assessment items
- Generate question banks for LMS import in JSON and human-readable Markdown
- Produce validated test sets when you need traceable source-based answers
Best practices
- Provide the exact filename when possible; if fuzzy, I will find the best match
- Prefer a source in .md; I will convert .pdf/.docx to Markdown when needed
- Specify total question count and any topic weight preferences up front
- Review the generated Markdown for pedagogical tone; use JSON for automated imports
- Keep original documents < 15,000 characters for single-mode generation or accept chunking for long texts
Example use cases
- Generate 20 questions from an annual report to assess comprehension of financials and strategy
- Convert course lecture notes into a 30-item practice quiz with mixed question types
- Create a certification-style exam by balancing easy/medium/hard items automatically
- Extract targeted review questions from a policy document for staff training
- Build a question bank from textbook chapters with per-chapter quotas in batch mode
FAQ
I accept Markdown directly. For .pdf, .doc, .docx and others I convert them to Markdown using markitdown after checking node/npm and required tools.
How do you ensure answers are correct?
I run a self-correction loop that fact-checks each question against the source, verifies logic and distractors, and validates JSON schema fields before output.
