A Claude Code skill that transforms scattered Computer Science & Technology classroom notes (Markdown format) into a structured, academically-deep knowledge base document through AI-powered knowledge expansion and systematic organization.
- Knowledge Expansion: Automatically identifies knowledge gaps (unexplained terms, missing derivations, assumed background) and fills them to academic-level completeness
- Knowledge Structuring: Organizes fragmented concepts into a hierarchical framework with prerequisite-ordered chapters and Mermaid concept relationship diagrams
- 9-Element Cognitive Lens System: Every expanded concept follows a fixed 5-element causal chain (问题 → 资源 → 抽象 → 机制 → 策略), with 4 analysis layers (权衡, 关系, 最小示例, 关键要点) embedded into the causal chain sections. 验证 is an optional standalone verification subsection. This is a cognitive model for reconstructing underlying mental models, not a rigid template — the goal is to reorganize knowledge by conceptual structure, not lecture order.
- Bilingual CS Terminology: English CS terms preserved with Chinese annotations on first use; code and formulas remain English-only
Computer Science & Technology only. Covers the following subfields:
| Category | Subfields |
|---|---|
| Theory | Algorithm Design & Analysis, Data Structures, Theory of Computation, Discrete Math, Mathematical Logic |
| Systems | Operating Systems, Computer Networks, Computer Architecture, Compilers, Distributed Systems |
| Data & AI | Database Systems, Artificial Intelligence, Machine Learning, Deep Learning, NLP, Computer Vision |
| Software | Programming Language Theory, Software Engineering, Design Patterns, Software Architecture |
| Security & Applications | Cryptography, Network Security, Computer Graphics, HCI |
Non-CS notes are rejected in Phase 1 by domain gating. Math/CS boundary domains (graph theory, information theory, probability theory) trigger a notice to confirm the CS-centric direction.
npx skills add makursi/note-refinernpx skills update note-refinerCheck CHANGELOG.md to see what's new before updating.
- Prepare your
.mdclassroom note files - In Claude Code, say: "refine my CS notes" (or reference specific file paths)
- The skill auto-loads and proceeds through 5 phases with minimal confirmation prompts:
| Phase | Name | Output |
|---|---|---|
| 1 | Ingest & Parse | Concept inventory, domain verification |
| 2 | Analyze & Identify Gaps | Gap analysis table (terminology audit + depth assessment) |
| 3 | Expand Knowledge | Academic-level expansions in 问题 → 资源 → 抽象 → 机制 → 策略 causal chain with embedded analysis layers (silent) |
| 4 | Structure Systematically | Hierarchical outline + Mermaid concept map |
| 5 | Generate Knowledge Base | Complete .md knowledge base document |
The output is saved to the source directory as [topic-slug]-knowledge-base.md.
- Gaps <= 8: confirms topic (Phase 1) and final outline (Phase 4) — 2 prompts
- Gaps > 8: adds gap priority confirmation (Phase 2) — 3 prompts
- User says "auto-proceed": only final outline shown for approval — 1 prompt
Generated knowledge bases contain 4 sections:
- Metadata header (date + source files only)
- Hierarchical table of contents
- Mermaid concept relationship diagram
- Main chapters (问题/资源/抽象/机制/策略 causal chain with embedded 权衡/关系/最小示例/关键要点, LaTeX formulas, formal definitions, comparison tables, glossary, optional verification code)
notes-refiner/
├── SKILL.md # Skill entry point (<=100 lines)
├── EXAMPLES.md # CS domain usage examples
├── CLAUDE.md # Claude Code collaboration guide
├── CHANGELOG.md # Version history
├── README.md # This file
└── references/
├── workflow.md # 5-phase detailed workflow instructions
├── output-schema.md # Output document template & formatting rules
└── expansion-guide.md # CS knowledge expansion methodology
- Chinese:
整理笔记知识扩展知识体系化笔记精炼课堂笔记 - English:
note refinementknowledge expansionknowledge structuringCS study notes
MIT