hwl668

hwl668 / Scientific-learning-skills-

Public

Diagnosis-first Agent Skills that turn AI from answer machine into learning tutor.

agent-skills ai-agent ai-agents ai-tutor claude-code
10
1
89% credibility
Found Jun 01, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

This project transforms AI assistants into teaching tutors that help you learn more effectively. Instead of dumping information, the AI first asks questions to understand exactly where you're confused, then explains only what you actually need. It includes 8 specialized learning skills for different situations — from first-time learning to mistake review to vocabulary memorization with spaced repetition. The system tracks what you struggle with and schedules reviews at the optimal time to help you remember. It's designed to work with popular AI coding assistants and study tools, making any AI a more effective teacher.

How It Works

1
📚 You want to really understand something, not just memorize it

Maybe you're stuck on a math concept, can't remember vocabulary, or keep making the same mistakes. You wish your AI helper would ask questions instead of just dumping answers.

2
⚙️ You install the learning skills to your AI assistant

A simple setup script creates a special connection that transforms your AI from an answer machine into a patient tutor who asks questions before explaining.

3
🤔 You ask a question naturally

You might say 'I can multiply matrices but don't get what they mean' or 'What's a limit? I'm learning this for the first time' — just like you'd ask a tutor.

4
🔍 The tutor diagnoses your exact confusion first

Instead of giving you everything, it asks a quick question: 'Did you learn this through equations or pictures?' Then it knows exactly what you're missing.

5
You get help matched to your situation
🌱
First time learning

Starts with a simple picture or analogy, builds intuition before formal definitions

🔮
Learned but fuzzy

Pinpoints the exact concept you're missing and explains just that gap

🎯
Can't solve a problem

Guides you step by step without spoiling the answer

📝
Word or text to memorize

Creates a memory plan and reminds you to review at the right time

6
✨ Your progress gets tracked automatically

Words you want to memorize are scheduled for review at optimal times. The tutor notices your weak spots and prioritizes them.

🎓 You understand deeply and remember what you learn

Because the tutor diagnosed your confusion, taught you the right way, and helped you practice at smart intervals, you actually get it — and you don't forget.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is Scientific-learning-skills-?

This is a Python framework that transforms AI assistants into diagnostic tutors rather than answer-dispensing machines. When you ask a question, the system first identifies your specific learning bottleneck before delivering targeted explanation. It provides nine specialized skills covering the full learning loop: zero-base explanations, fuzzy concept diagnosis, problem-solving guidance, mistake review, word etymology deep-dives, text memorization with spaced repetition, and study plan building.

The framework includes a rule-based router that classifies questions to the appropriate skill, a cognitive diagnosis module detecting six types of learning bottlenecks, and an SM-2 style memory scheduler for tracking review intervals. A learned router using TF-IDF character n-grams with logistic regression provides an upgrade path.

Why is it gaining traction?

The hook is the diagnosis-first philosophy. Instead of dumping encyclopedic responses, the system asks clarifying questions first to pinpoint exactly where you're stuck. The README demonstrates this clearly: asking about matrix multiplication gets a response that first diagnoses whether your issue is intuition versus calculation, then addresses only what you're missing.

The framework is platform-agnostic. Skills are Markdown instruction sets that work with Claude Code, OpenAI Codex, OpenClaw, or any agent via system prompt injection. Token budgeting lets you deploy subsets of the nine skills when context is limited.

Who should use this?

Developers building AI-powered tutoring systems or educational tools will find the routing and evaluation infrastructure valuable. Learners studying technical subjects who find AI responses overwhelming will benefit from the diagnosis-first approach. Teams wanting to standardize AI-assisted learning across different agent platforms can use the platform-agnostic skill definitions.

Verdict

At 10 stars with a credibility score of 0.90%, this is an early-stage project with a compelling conceptual foundation but limited community validation. The code quality appears solid with testable evaluators and a clean module structure, but the low star count means you're an early adopter bearing typical early-stage risks. Worth exploring for the learning design ideas, but don't bet production systems on it yet.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.