Skip to content

sjquant/cram

v0.1.0MIT

Turn any document into an interactive flashcard quiz in a single self-contained HTML file. Extracts normalized cards — basic, multiple choice and cloze — then renders an offline quiz that remembers your progress and lets you retry only what you missed.

Cram

English | 한국어

Turn anything into a quiz.

Reading something isn't the same as knowing it. Cram turns your notes, documents, and web material into flashcards so you can test your understanding without writing questions yourself.

Create a quiz with your AI agent, review what you missed, and keep it as a single HTML file you can study offline or share.

Try the live demo → — no install, nothing to clone. Make your own quiz ↓

Cram walkthrough: request flashcards, then reveal answers and review a quiz

Explore the demo's deck source or open its HTML file from a local clone.

Why Cram?

  1. Spend your time studying, not making flashcards. Give your AI agent material you want to learn. The Cram skill guides it through creating questions and rendering a quiz you can use right away.
  2. Focus your next round on what you missed. Test yourself, reveal answers, and grade your recall. Retry just the missed cards, and come back later with your place and results saved in your browser when storage is available.
  3. Keep the quiz. Share it. Study offline. Each quiz is one HTML file with the cards and player included. Open it or send it to someone else—playing it needs only a browser, with no account, app installation, or server.

Make your first quiz

You'll need a supported AI agent and system python3 to create a quiz. The renderer requires no additional Python packages.

1. Install the skill

In Claude Code, add this repository as a community plugin source, then install Cram:

/plugin marketplace add sjquant/cram
/plugin install cram@cram

This installs directly from sjquant/cram, not an official plugin catalog.

For other AI agents, run this in your terminal and select your agent:

npx skills add sjquant/cram --skill cram

This installation route uses Node.js/npm to run npx. More installation options ↓

2. Give your agent something to study

Save your notes as notes.md in your working directory, then ask Claude Code:

/cram:cram Read ./notes.md and create 10 flashcards to test the key concepts.
Save the quiz as study.html.

Claude Code prefixes plugin skills with the plugin name, so the command is /cram:cram. If you installed with npx skills instead, it's /cram.

In another agent, ask: “Use the cram skill to read ./notes.md and create 10 flashcards to test the key concepts. Save the quiz as study.html.” You can also paste material, attach a document, or provide a web URL your agent can access.

3. Open the quiz

Open the generated study.html in your browser. Work through the cards, reveal the answers, and grade yourself. At the end, review your results or retry only the cards you missed.

You can reopen the file offline or send it to someone else. Your study progress stays in your browser; it isn't included when you share the file.

Want to send a link instead? Share your quiz with free hosting →

What could you study?

  • Technical documentation: turn a guide into questions that check whether you can explain its concepts. Try the HTTP caching demo (source).
  • Language study: drill the mistakes learners actually make. 한국인이 자주 틀리는 일본어 (source) covers the Japanese traps Korean speakers fall into, written natively in Korean.
  • Fandom trivia: quiz your friends on what they love. K-pop Trivia (source) covers hit songs, fandoms, fanchants, and memes.
  • Course notes: make a quiz from a lecture before your next study session.
  • Onboarding material: create a quiz from a team guide and share the HTML with new teammates.

Study your way

Practice with basic question-and-answer cards, multiple-choice questions, and fill-in-the-blank cards. Decks can include hints and explanations. The player supports keyboard navigation, reduced-motion preferences, and high-contrast mode.

Study order, saved progress, and keyboard settings

In Settings → Study order, Shuffle starts a new round with hidden answers while keeping recorded grades in learning history. Restore original order returns to the source order. Retry rounds keep their missed-card scope.

Session order, position, results, and Cram-mode repeat attempts are saved in the browser when storage is available. Saved sessions belong to the same deck in the same browser; changing the deck's card content starts a new session. To start over partway through, use Reset progress in Settings; after a warning, it clears that deck's answers and place while keeping your settings.

Settings lets you turn off the A/H single-letter shortcuts; the choice is remembered when browser storage is available.

Player language

Study with controls in English, Korean, Japanese, Simplified Chinese, Spanish, or French. Ask your agent for the player language you want. Card language is separate: ask for translated cards if you want those changed too.

More installation options

Choose agents with the skills installer

The npx skills installer places the skill in a supported agent's skill directory. To target agents explicitly, repeat --agent, for example:

npx skills add sjquant/cram --skill cram \
  --agent codex \
  --agent cursor \
  --agent github-copilot \
  --agent grok \
  --agent kiro-cli \
  --agent antigravity-cli

The portable skill lives in skills/cram/. For the extraction, deck-format, and rendering workflow, see the skill guide.

Requirements and scope

  • Creating a quiz: requires an AI agent that can read your source material and run the Python renderer. Source access, account requirements, costs, and data handling depend on the agent and model you use.
  • Playing a quiz: requires only a browser. The generated HTML includes the player and deck data, with no server or network connection needed.
  • Validation: Cram checks the deck's structure before rendering. This does not fact-check AI-generated questions or answers; review them against your source material.

License

MIT