Getting started

1. Start

Install Claude Code if you haven't. Then:

git clone <this repo> your-ai-tutor && cd your-ai-tutor
claude

Say I want to learn <topic>. The tutor asks five short questions (what you want to be able to do, what you already know, how you like to be taught, which days you can show up, and whether you want questions, a project, or a coach beside you) and creates envs/<topic>/. It ends by telling you the phrase to start: let's do <topic>.

The skills load from .claude/skills the moment claude starts in this folder. There is no install step.

2. What it made

Everything about your topic is plain text in envs/<topic>/. Edit any of it.

• env.toml: tier, which days you show up, how strict the tutor is.
• CLAUDE.md: what you're learning, what you already know, what "done" means.
• roadmap.md: 5–10 units, each with a "done when" sentence.
• tutor.md: how you want to be taught: voice, how questions are asked, analogies to use or avoid, whether you want diagrams. Add to it any time by saying "remember I prefer …". It shapes how the tutor teaches, never whether it hands you the answer; that stays in env.toml.

Prefer to start from a template? Copy one from examples/envs/ (practice-drills, project, creative-draft, language-practice, math-science, _template) into envs/<name>/ and set name in its env.toml.

3. A session

Say	Effect
let's do <topic>	Start or continue the env
what should I do today	Name today's block from your envs' cadences
pause / checkpoint	Save your spot
status	One-line readout
wrap up	Save, journal, tick the swim lane
quiz me / test me	A grill on what you have covered so far
retro	Debrief and queue improvements

4. Env settings that matter

Setting	Try this when
terminal = "forbidden"	You want zero chance of the agent running your code
register = "plain"	The topic is new to you; questions get simpler
mode = "project"	You are making a real artifact
tier = "sprint" + expires	A deadline is coming; it outranks everything until that date

5. Optional extras

Each one is off until you want it, and the skill tells you what to set up the first time you ask.

• Flashcards: say card this during a session. Needs Anki desktop with the AnkiConnect add-on; see harness/plugins/anki/.
• Reverse engineering a live site: say reverse engineer <url>. Needs the Claude-in-Chrome extension; see REVERSE-ENGINEERING.md.
• Study material from PDFs or video: say extract this pdf. Needs pdftotext and ffmpeg.
• Checking the tutor itself: python3 -m your_ai_tutor.cli.evals run --rule route runs a few scripted requests and reports whether each reached the right skill. It spends model calls; see the README's contributor section.
• The tutor in other projects: python3 -m your_ai_tutor.cli.install registers the harness as a Claude Code plugin. Inside this repo it is not needed.

Your data

envs/, media/, tracker/ and memory/ are git-ignored so personal progress never lands in a public fork by accident. Remove the lines at the bottom of .gitignore if you want to version them in a private repo.