PromptEducation & Research
Plan a study session from supplied material
Create explanations and practice questions bounded by a learner's goal and provided sources.
✳Claude
By Dave AIDave AI original
MIT · FreeView prompt
Create effective flashcards with optimal formatting, spaced repetition integration, and memory science principles.
---
name: flashcards
slug: flashcards
version: 1.0.0
description: Create effective flashcards with optimal formatting, spaced repetition integration, and memory science principles.
homepage: https://clawic.com/skills/flashcards
metadata:
clawdbot:
emoji: 🃏
displayName: Flashcards
---
## Card Formulation Rules
**One fact per card**: Never combine multiple concepts. "What is X?" not "What are X, Y, and Z?"
**Atomic questions**: Break complex topics into smallest testable units. Each card tests exactly one thing.
**Bidirectional cards for definitions**: Create both term→definition AND definition→term to prevent recognition-only learning.
**Use cloze deletions for facts**: "The mitochondria is the {{c1::powerhouse}} of the cell" forces active recall.
## Question Types by Effectiveness
**Best retention**: Why/How questions that require understanding, not just recall.
**Good retention**: Fill-in-the-blank, definition recall, process steps.
**Weak retention**: Yes/No questions, multiple choice (use sparingly).
**Avoid**: Questions answerable by pattern matching or elimination.
## Anki-Specific Formatting
**TSV import format**: `front\tback\ttag1 tag2` — tabs separate fields, spaces separate tags.
**Cloze syntax**: `{{c1::answer}}` for single deletion, `{{c1::first}} and {{c2::second}}` for multiple.
**Image occlusion**: Use for diagrams, maps, anatomical images. Hide labels, reveal on flip.
**Tags for organization**: Use hierarchical tags `subject::topic::subtopic` for filtered study.
## Memory Science Integration
**Minimum information principle**: Simpler cards = better retention. If card feels complex, split it.
**Personal connection**: Add context from your experience. "X reminds me of Y" strengthens encoding.
**Concrete over abstract**: "Paris is capital of France" beats "Capitals are important cities."
**Imagery when possible**: Visual descriptions enhance memory. "Mitochondria = bean-shaped power plant."
## Common Mistakes
**Too much text on back**: Keep answers under 20 words. Long answers = weak recall signal.
**Orphan cards**: Cards without context fail. Include source/chapter in tags.
**Copy-paste from textbook**: Rephrase in your own words. Understanding before memorization.
**Skipping hard cards**: Difficulty means you need it most. Never suspend without replacement.
## Output Formats
**Anki TSV**: `question\tanswer\ttag1 tag2`
**Quizlet import**: Question and answer separated by tab, cards separated by newline.
**Markdown table**: For review before import.
```
| Front | Back | Tags |
|-------|------|------|
| Q1 | A1 | topic |
```
## Spaced Repetition Settings
**New cards/day**: 10-20 for sustainable learning. More causes review pile-up.
**Review intervals**: Trust the algorithm. Don't manually reschedule.
**Again vs Hard**: "Again" = complete failure (resets interval). "Hard" = struggle but recalled.
**Leeches**: Cards failed 8+ times need rewriting, not more repetition.
---
name: flashcards
slug: flashcards
version: 1.0.0
description: Create effective flashcards with optimal formatting, spaced repetition integration, and memory science principles.
homepage: https://clawic.com/skills/flashcards
metadata:
clawdbot:
emoji: 🃏
displayName: Flashcards
---
## Card Formulation Rules
**One fact per card**: Never combine multiple concepts. "What is X?" not "What are X, Y, and Z?"
**Atomic questions**: Break complex topics into smallest testable units. Each card tests exactly one thing.
**Bidirectional cards for definitions**: Create both term→definition AND definition→term to prevent recognition-only learning.
**Use cloze deletions for facts**: "The mitochondria is the {{c1::powerhouse}} of the cell" forces active recall.
## Question Types by Effectiveness
**Best retention**: Why/How questions that require understanding, not just recall.
**Good retention**: Fill-in-the-blank, definition recall, process steps.
**Weak retention**: Yes/No questions, multiple choice (use sparingly).
**Avoid**: Questions answerable by pattern matching or elimination.
## Anki-Specific Formatting
**TSV import format**: `front\tback\ttag1 tag2` — tabs separate fields, spaces separate tags.
**Cloze syntax**: `{{c1::answer}}` for single deletion, `{{c1::first}} and {{c2::second}}` for multiple.
**Image occlusion**: Use for diagrams, maps, anatomical images. Hide labels, reveal on flip.
**Tags for organization**: Use hierarchical tags `subject::topic::subtopic` for filtered study.
## Memory Science Integration
**Minimum information principle**: Simpler cards = better retention. If card feels complex, split it.
**Personal connection**: Add context from your experience. "X reminds me of Y" strengthens encoding.
**Concrete over abstract**: "Paris is capital of France" beats "Capitals are important cities."
**Imagery when possible**: Visual descriptions enhance memory. "Mitochondria = bean-shaped power plant."
## Common Mistakes
**Too much text on back**: Keep answers under 20 words. Long answers = weak recall signal.
**Orphan cards**: Cards without context fail. Include source/chapter in tags.
**Copy-paste from textbook**: Rephrase in your own words. Understanding before memorization.
**Skipping hard cards**: Difficulty means you need it most. Never suspend without replacement.
## Output Formats
**Anki TSV**: `question\tanswer\ttag1 tag2`
**Quizlet import**: Question and answer separated by tab, cards separated by newline.
**Markdown table**: For review before import.
```
| Front | Back | Tags |
|-------|------|------|
| Q1 | A1 | topic |
```
## Spaced Repetition Settings
**New cards/day**: 10-20 for sustainable learning. More causes review pile-up.
**Review intervals**: Trust the algorithm. Don't manually reschedule.
**Again vs Hard**: "Again" = complete failure (resets interval). "Hard" = struggle but recalled.
**Leeches**: Cards failed 8+ times need rewriting, not more repetition.
Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.