generic-tutor

A generic AI tutor skill using SM-2 spaced repetition. File-based state, markdown knowledge base, no dependencies beyond bash and jq.

How It Works

1. An AI agent loads SKILL.md to become a tutor
2. Knowledge lives in knowledge/concepts/ as markdown files
3. Learner state is tracked in state/{topic}.json
4. SM-2 algorithm schedules reviews automatically

Quick Start

# Initialize a topic
bash scripts/init.sh git-basics

# Check what's due
bash scripts/due.sh git-basics

# View progress
bash scripts/stats.sh git-basics

# Then: load SKILL.md in an AI agent session and say "start a session on git-basics"

Adding a New Topic

1. Create knowledge/concepts/ files using templates/concept-template.md
2. Create knowledge/manifest.json using templates/manifest-template.json
3. Run bash scripts/init.sh {topic}

Structure

SKILL.md              # Agent loads this to become a tutor
config.json           # Learner config
knowledge/            # Markdown knowledge base
  manifest.json       # Concept index + prerequisite DAG
  concepts/           # One file per concept
state/                # Learner progress (git-ignored)
scripts/              # CLI helpers
templates/            # Starting points for new topics

Grading (0-5)

Grade	Meaning
5	Perfect recall
4	Correct with minor hesitation
3	Correct but effortful
2	Wrong but close
1	Minimal relevant knowledge
0	Complete blank