learn-skill
guepardlover77/learn-skill
A Claude Code skill that turns any subject into a science-backed learning system — Anki cards, Obsidian notes, and spaced repetition in one command.
Instalação
npx convez-ai guepardlover77/learn-skill1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
learn-skill
A Claude Code skill that turns any subject into a science-backed learning system — Anki cards, Obsidian notes, and a spaced repetition schedule, in one command.
What it does
You give Claude any learning material — a PDF, a concept, a vocabulary list — and learn-skill builds a complete learning system around it:
| Output | Description |
|---|---|
| Adaptive learning plan | Detects content type, asks 2-3 targeted questions, generates a session plan calibrated to your available time |
| Anki deck (.apkg) | Exports a ready-to-import Anki file with Basic, Cloze, and reverse cards — no manual work |
| Obsidian note | Structured note with concept map, active recall checklist, Feynman section, and review schedule |
| Spaced repetition schedule | SM-2 review dates (J+1, J+6, J+14, J+30, J+60) baked into the Obsidian note |
Works for: courses & text, abstract concepts, foreign languages — in any language.
Demo
/learn-anything
> I pasted my thermodynamics chapter (12 pages)
Claude detects: Type A — Text/Course
Questions:
1. Objectif : comprendre en profondeur ou mémoriser ? → Comprendre
2. Temps disponible aujourd'hui ? → 1h30
3. Flashcards Anki ? → Oui
Generating...
✓ Concept map (8 core concepts)
✓ Full session plan (3 × 50 min Pomodoros)
✓ Feynman checkpoint for entropy and enthalpy
✓ 34 Anki cards → thermodynamics.apkg
✓ Obsidian note → thermodynamics.md (with review schedule)
Next review: Tomorrow, 15 min active recall only.
Science
This skill implements five evidence-based learning techniques:
1. Spaced Repetition (SM-2 algorithm)
Each card gets review intervals that expand as recall improves: 1d → 6d → 14d → 30d → 60d. The ease factor (default 2.5) adjusts per card based on recall quality.
2. Active Recall (Testing Effect)
Every session ends with retrieval practice — no passive re-reading. Retrieving information strengthens the memory trace 2–3× more than reviewing the source material.
3. Feynman Technique
After encoding, Claude prompts you to explain the concept in plain language. The points where you struggle are your exact knowledge gaps — those gaps generate targeted Anki cards.
4. Chunking
Content is broken into groups of 4±1 concepts (Cowan's updated model of working memory), reducing cognitive load during encoding.
5. Desirable Difficulties
The skill forces the generation effect: you attempt to recall before seeing the answer. This harder retrieval creates stronger, more durable encoding.
Installation
curl -fsSL https://raw.githubusercontent.com/guepardlover77/learn-skill/main/install.sh | bash
Or manually:
git clone https://github.com/guepardlover77/learn-skill
cd learn-skill
./install.sh
Then restart Claude Code. The skill activates automatically when you say "I want to learn X" or via /learn-anything.
Usage
Trigger phrases (any language)
/learn-anything- "Je veux apprendre [X]"
- "Help me memorize / understand / study [X]"
- "Generate Anki cards for [X]"
Workflow
- Paste your material (text, notes, vocabulary list) or describe a concept
- Answer 2-3 questions (goal, time available, Anki yes/no)
- Get your learning system: plan + Anki script + Obsidian note
- Run the Python script to export your
.apkgAnki deck - Follow the review schedule — Claude tracks it in Obsidian
Examples
See examples/ for real session outputs:
physics-thermodynamics.md— 12-page course → 34 Anki cardsjapanese-n5-vocab.md— 50 words → bidirectional deckquantum-computing.md— abstract concept → concept map + Feynman
Requirements
- Claude Code (any plan)
- Python 3.8+ with
genankifor Anki export:pip install genanki - Obsidian (optional, for progress tracking)
- Anki (optional, for spaced repetition)
How it works
The skill is a single SKILL.md file that Claude Code reads as instructions. It contains:
- A content-type detector (text/course vs abstract concept vs language)
- An adaptive interview (2-3 questions max, tuned per content type)
- A session planner (sprint / full / deep-dive based on available time)
- A card generator (Basic, Cloze, Reverse — genanki-compatible Python output)
- An Obsidian template (concept map, Feynman section, review log)
- A post-session protocol (what to do in the next 24h)
Contributing
PRs welcome. Especially interested in:
- New content-type detectors (e.g. code/programming, music theory)
- Better card templates for specific domains
- Anki-Connect integration (auto-import without the Python script)
- MCP server version
License
MIT — use it, fork it, improve it.