AgentSkillsCN

notebook-patterns

为丰富笔记本内容而设计的教学模式(如博弈论模型)。适用于在Jupyter笔记本中添加解读、梳理笔记结构,或创作教育性内容的场景。

SKILL.md
--- frontmatter
name: notebook-patterns
description: Pedagogical patterns for enriching notebooks (GameTheory model). Use when adding interpretations, structuring notebooks, or creating educational content in Jupyter notebooks.

Notebook Enrichment Patterns

Standard Header (all notebooks)

markdown
# Series-N-Title

**Navigation** : [Index](README.md) | [<< Precedent](Series-N-1.ipynb) | [Suivant >>](Series-N+1.ipynb)

## Objectifs d'apprentissage

A la fin de ce notebook, vous saurez :
1. ...
2. ...

### Prerequis
- Python 3.10+ / .NET 9.0
- Cle API configuree (.env)

### Duree estimee : XX minutes

Interpretation Pattern (after significant code output)

markdown
### Interpretation : [Concept Name]

**Sortie obtenue** : [Brief summary of output]

| Aspect | Valeur | Signification |
|--------|--------|---------------|
| ... | ... | ... |

**Points cles** :
1. ...
2. ...

> **Note technique** : [Detail if relevant]

Positioning Rules (CRITICAL)

  • Introduction cells go BEFORE the code they introduce (future tense)
  • Interpretation cells go AFTER the code output they analyze (past tense)
  • Transition cells go BETWEEN sections
  • Conclusion cells go at the END of a section or notebook

Cell Insertion Strategy

  1. Work from BOTTOM to TOP to avoid index shifting
  2. Use cell_id (not index) for NotebookEdit insert reference
  3. Verify position immediately after each insertion
  4. One cell at a time - never batch inserts

Domain-Specific Vocabulary

DomainKey Terms
MLaccuracy, loss, epoch, overfitting, cross-validation
Probas/Inferprior, posterior, distribution, inference, factor graph
GameTheoryNash equilibrium, Shapley value, dominant strategy
SymbolicAIsatisfiability, resolution, proof, axiom, theorem
GenAIprompt, tokens, embedding, fine-tuning, hallucination
Optimizationfitness, generation, mutation, crossover, convergence

Code/Markdown Ratios

LevelCodeMarkdownVisualizations
Intro35-40%55-60%min 3
Intermediate45-50%45-50%min 4
Advanced55-60%35-40%min 2

Quality Checklist

  • Navigation header present
  • Learning objectives stated
  • No consecutive code cells without explanation
  • Every significant output has interpretation
  • Summary table at end of each major section
  • Conclusion with takeaways
  • Progressive difficulty (simple to complex)