Installable skillApache-2.0

Anki Card Generator

生成高质量 Anki 闪卡,遵循原子化原则和认知科学最佳实践。输出 simple-anki-sync 兼容的 markdown 表格。当用户说「生成 Anki 卡片」「做卡片」「帮我记忆」「做闪卡」「学习卡片」「背诵卡」「记忆这个」时触发。

BY David-Li0406UPDATED Oct 27, 2026
INTENDED FOR◈OpenClawInstall in a listed agent.

instructions.md

---
name: anki-card-generator
description: 生成高质量 Anki 闪卡,遵循原子化原则和认知科学最佳实践。输出 simple-anki-sync 兼容的 markdown 表格。当用户说「生成 Anki 卡片」「做卡片」「帮我记忆」「做闪卡」「学习卡片」「背诵卡」「记忆这个」时触发。
---

# Anki Card Generator

Generate high-quality Anki cards based on LessWrong best practices and simple-anki-sync format.

## Pre-Generation Clarification

Before generating cards, clarify with user if unclear:

1. **Scope**: "Which aspects to focus on?" (for broad topics)
2. **Depth**: "Basic concepts only, or detailed Level 2/3 cards?"
3. **Quantity**: "How many cards? (Recommend 5-10 for core concepts)"
4. **Context**: "Any specific exam or application scenario?"

Proceed only after understanding requirements.

## Output Format

Use simple-anki-sync format:

```markdown
#anki/[domain]/[topic]

| [Question] |
| ---------- |
| [Core answer]<br><br><small>💡 [Supplementary info]</small> |
```

### Format Options

**Option A (Recommended)**: HTML tags

```markdown
| 唐朝建立时间 |
| ---------- |
| 618年,李渊建立<br><br><small>💡 隋末农民起义后起兵</small> |
```

**Option B**: Separator

```markdown
| 唐朝建立时间 |
| ---------- |
| 618年,李渊建立 ——— 💡 隋末农民起义后起兵 |
```

**Option C**: Parentheses

```markdown
| 唐朝建立时间 |
| ---------- |
| 618年,李渊建立(趣闻:其子李世民功劳最大) |
```

## Atomization Rules

### Word Limits

- **English**: Max 9 words, absolute limit 18 words
- **Chinese**: Recommended 15-20 characters, absolute limit 30-35 characters
- **Max items**: 3 bullet points per card

### Core Principle

If a card can be split into two shorter cards, split it.

## Question Design

### Standardized Templates

- **Time**: "X 时间" (not "X发生于何时?")
- **Definition**: "X 定义" (not "什么是X?")
- **Person**: "who X" (not "谁做了X?")
- **Pros/Cons**: "X 利弊" (not "X的优势是什么?")

### Key Rules

- Match real-world recall scenarios
- Use plain, unremarkable wording
- Avoid words in question that appear in answer
- Keep all critical info in answer, not question

## Answer Construction

### Core Answer

- Strictly follow word limits
- Answer should be meaningful without the question
- All key information in answer

### Supplementary Info (Optional)

Format: `<br><br><small>💡 content</small>`

**Emoji Guide**:

- 💡 Fun fact / trivia
- 📝 Note / explanation
- 🔗 Related concept
- ⚡ Tip / key point
- 📊 Data / statistics
- 📅 Date / timeline

Keep supplementary info to 10-20 characters.

### Handle System

Use `>` to reference related cards:

```markdown
| 牛顿贡献 |
| ------- |
| >运动定律 >万有引力 >微积分发展 |
```

## Detail Levels

- **Level 1**: Basic concept (core answer)
- **Level 2**: Detailed info (supplementary section)
- **Level 3**: Advanced details (create separate cards)

## Tag Naming

Use English tags: `#anki/[domain]/[topic]`

Common domains: history, programming, language, science, mathematics, psychology, economics, philosophy, medicine, art

## Quality Checklist

### Per Card

- [ ] Core answer within word limit?
- [ ] Correct supplementary format?
- [ ] Can it be further split?
- [ ] Question matches real recall scenario?
- [ ] No memory shortcuts?
- [ ] All key info in answer?

### Card Set

- [ ] Appropriate cross-references?
- [ ] Proper detail levels?
- [ ] No redundancy?

## Domain Examples

See [references/examples.md](references/examples.md) for detailed examples:

- History (ancient China)
- Programming (Python)
- Language learning (English)
- Academic concepts (psychology)
- Complex topics (quantum mechanics)
- Error corrections
- Advanced techniques (reversible cards, redundancy design)

## Workflow

1. **Receive input**: Knowledge points, wiki links, study materials
2. **Clarify**: Ask questions if uncertain
3. **Generate**: Follow atomization and best practices
4. **Output**: simple-anki-sync markdown format

**Core Philosophy**: Prioritize sustainability and real recall scenarios over comprehensive coverage. Focus on preventing cognitive and motivational barriers.
COMPATIBILITY & REFERENCES

Platform guidance

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.