A Claude skill for generating atomic, cognitively-principled flashcards from technical source material.
Upload lecture notes, textbooks, or technical documentation. Get interactive flashcards organised into three cognitive layers:
| Layer | Tests | Example |
|---|---|---|
| L1: Recall | Facts, definitions, formulas | "Define the Jacobian matrix" |
| L2: Understanding | Why/how, intuitions | "Why is the Jacobian useful for coordinate transforms?" |
| L3: Boundaries | Limitations, edge cases | "When does the Jacobian become singular?" |
The skill enforces atomicity (one concept per card), refuses to card inappropriate content (proofs, worked examples), and renders mathematical notation via KaTeX.
- Clone this repo or download as ZIP:
git clone https://github.com/jalliet/flashcards.git - Create a skill package:
zip -r flashcards.zip SKILL.md references/ assets/ - Upload to Claude.ai: Settings > Capabilities > Skills > Upload Skill
Upload a PDF, paste lecture notes, or describe a topic:
"Generate flashcards spanning topicas from all the project files."
"Create flashcards covering gradient descent, including failure modes"
"Make revision cards for chapters 3-5 of the uploaded PDF"
"Generate only L3 (boundary) cards for this material"
"Focus on L1 recall cards for definitions and formulas"
"Card 12's formula is missing the inverse; fix it"
"Split card 7 into two separate cards"
"Add more L2 cards for the eigenvalue section"
Create a Claude Project for your course/subject. Add:
- This skill (
.skillfile) - Your lecture notes and readings
- Project instructions like: "Always use the flashcard skill when I ask for revision materials"
For best results with complex documents, also install from anthropics/skills:
| Skill | Why |
|---|---|
pdf |
Better extraction from scanned/complex PDFs |
docx |
Preserves formatting from Word documents |
frontend-design |
Improves React artifact rendering and styling for academic content |
The flashcard skill works without these, but they improve source parsing.
The skill produces an interactive React artifact with:
- Layer filtering (L1/L2/L3 tabs)
- Topic filtering
- Star/favourite cards
- Shuffle mode
- Responsive design
Cards can be copied to Anki or other spaced repetition systems.
generating-stem-flashcards/
├── README.md # This file
├── THEORY.md # Scientific foundations (Bloom, CLT, etc.)
├── SKILL.md # Main skill instructions
├── references/
│ ├── COGNITIVE_LAYERS.md # L1/L2/L3 definitions
│ ├── ATOMICITY.md # Quality rules, refusal policy
│ └── LATEX_SYNTAX.md # KaTeX reference
└── assets/
└── flashcard_template.jsx
Note: Only the skill files (SKILL.md, references/, assets/) are packaged in the .skill file. README.md and THEORY.md are GitHub documentation only.
Most AI flashcard generators produce low-quality cards: compound questions, no cognitive framework, no quality control.
This skill encodes learning science directly into generation:
- Bloom's Taxonomy → Three-layer structure
- Cognitive Load Theory → Atomicity rules
- Minimum Information Principle → Refusal policy
See THEORY.md for the full scientific foundation with citations.
MIT