Universal Study Material Generator

Behind the Story

My daughter attends Contour Education, a tutoring institution in the Melbourne CBD. Every session was close to a full Sunday: 2-3 hours in the classroom, plus roughly 4 hours round-trip by train. And it wasn’t cheap — about $1,200 per subject, per term of sessions. Both the money and the time were adding up.

That cost made me ask the obvious question: what is Contour actually selling? It isn’t the curriculum — that’s in any textbook, freely available. So what’s the part worth $1,200 and a lost Sunday?

The answer turned out to be the question style, not the content. A standard textbook asks “find the area of this rectangle.” A Contour test asks something like “you’re building a tennis court of this size — how much land do you need, accounting for the surrounds?” Same underlying maths, but transformed into a realistic, applied scenario that actually tests whether a student understands the concept well enough to use it, not just recall it.

That’s the premise this project tests: if the expensive part is the question-transformation pattern, not the raw curriculum, then an AI system that can learn both — the content (what’s being taught) and the pattern (how it gets turned into a realistic, applied question) — from real study materials should be able to reproduce that value. This system was built to test exactly that: give it a workbook (or a textbook scan) plus a sample of Contour-style questions, and see whether it can learn the pattern well enough to generate fresh, realistic questions on new material — not just reworded textbook questions.

Transform any educational content into unlimited fresh practice materials. This system works for ANY subject—Mathematics, IELTS, Science, History, Programming, and more.

Architecture update: This post originally described two separate systems. The project has since consolidated onto one unified 3-agent pipeline — see below.

How It Works

The system doesn’t “know” any subject—it LEARNS from whatever materials you provide and generates fresh, curriculum-aligned content.

Two scenarios depending on what you have:

Choose Your Scenario

Do you have sample tests/exercises?
    │
    ├─ YES → Scenario A: Workbook Input
    │         (Workbook PDF + Test PDF)
    │
    └─ NO  → Scenario B: Textbook Scan
              (3-section scan: Summary, Checklist, Review)

Both scenarios feed the SAME 3-agent pipeline:
curriculum-analyzer → test-contents-extractor → curriculum-aligned-question-generator

Scenario Comparison

Aspect	Scenario A: Workbook	Scenario B: Textbook Scan
You Have	Curriculum + Test samples	Curriculum only
Input	Workbook PDF + Test PDF	Textbook scan (Summary, Checklist, Review sections)
How It Works	curriculum-analyzer learns from your workbook + test PDF	curriculum-analyzer learns from the 3-section textbook scan instead
Output	Variations of existing tests	Brand new tests from scratch
Best For	Creating practice versions	New subjects without test samples
Architecture	Same unified 3-agent pipeline (curriculum-analyzer → test-contents-extractor → curriculum-aligned-question-generator)	Same unified 3-agent pipeline (curriculum-analyzer → test-contents-extractor → curriculum-aligned-question-generator)

Only the input to the first agent, curriculum-analyzer, changes between scenarios. Everything downstream — test-contents-extractor and curriculum-aligned-question-generator — is identical.

Subject Examples

Subject	Scenario A (with tests)	Scenario B (curriculum only)
Mathematics	Vary existing mock tests	Create tests from textbook chapters
IELTS/TOEFL	Vary Cambridge practice tests	Create from grammar/vocab books
Science	Vary lab exams	Create from textbook review sections
History	Vary past papers	Create from chapter summaries
Programming	Vary coding challenges	Create from tutorial exercises
Medical/Nursing	Vary board exam questions	Create from study materials
Music Theory	Vary grade exam papers	Create from theory textbooks
Law	Vary bar prep questions	Create from casebooks

Key Capabilities (Both Scenarios)

Fresh Content — Genuinely new questions, not just number changes
Curriculum-Aligned — Stays within scope of provided materials
Pattern Library — 49+ question formats that grow with each test
Diagram Support — Generates new visual content automatically
PDF Output — Professional, print-ready format
Answer Sheets — Complete solutions with worked steps

Who Can Use This

Teachers — Create unlimited practice tests for any subject
Students — Get endless fresh practice without repeating questions
Test Prep Companies — Scale content creation massively
Parents — Help children study any subject at home
Corporate Training — Generate assessment materials for employees

Detailed Architecture

For technical details on how the pipeline works:

Curriculum Test Generator Architecture Diagrams — The unified 3-agent pipeline: curriculum-analyzer, test-contents-extractor, and curriculum-aligned-question-generator, run sequentially with resume logic
Contour Test Generator Skill Architecture — Historical reference only; describes an earlier, now-superseded design

Explore the Documentation

Curriculum Test Generator – Architecture Diagrams — the current 3-agent pipeline in full technical detail (curriculum-analyzer, test-contents-extractor, curriculum-aligned-question-generator).
Contour Test Generator Skill – Architecture (historical) — an earlier knowledge-base/LaTeX approach that was tried and abandoned in favor of the unified pipeline above. Kept for reference only.

Repository: github.com/strider73/contour_questionbook_creator

The Vision

This system proves that AI can learn any curriculum and generate high-quality practice materials. Whether it’s your daughter’s Year 10 math test or your own IELTS preparation—the same architecture handles it all.

The only requirement: provide the source materials, and the system learns what to teach and how to test it.