zhongwei/gh-junghan0611-org-mode-skills-denote-org

Files

Zhongwei Li f72e181b3c Initial commit

2025-11-30 08:29:49 +08:00

11 KiB

Raw Blame History

Denote-Org Skills for Claude

Bringing Anthropic's Life Sciences paradigm to Life Everything

Comprehensive Denote PKM system support for Claude AI, validated with 3,000+ org files in production knowledge bases.

Denote-first. Org-mode powered. Life-scale proven.

🌟 Vision: Life Sciences → Life Everything

Anthropic proved that domain context + AI = expert-level collaboration with Life Sciences (PubMed, Benchling, 10x Genomics).

This project extends that paradigm from Biology to Living:

Life Sciences (Biology) → Life Everything (Living)

Just as Claude becomes a scientist with PubMed context, Claude becomes your personal knowledge partner with Denote context.

The Pattern

Domain	Context	Result
Anthropic	PubMed (millions of papers) + Claude	Scientific AI
This Project	Denote (3,000+ org files) + Claude	Personal Knowledge AI

Same methodology. Different domain. Universal paradigm.

Why This is a Skill, Not Just a Prompt

Anthropic's Skills Philosophy:

"Skills transform Claude from a general-purpose agent into a specialized agent equipped with procedural knowledge that no model can fully possess."

— Equipping agents for the real world with Agent Skills

Just like:

PDF Skill solves complex forms (8 Python scripts for bounding boxes, validation...)
XLSX Skill maintains formulas (recalc.py with LibreOffice integration)
Denote-Org Skill handles 3,000-file knowledge graphs (finder, graph builder, silo manager)

Skills are NOT prompts. They are operational systems.

The Problem This Solves

Without This Skill

❌ Claude doesn't know Denote file naming: 20251021T105353--title__tags.org ❌ Finding files in 3,000+ requires repeated token-heavy searches ❌ Knowledge graph navigation is slow and error-prone ❌ Literate programming (:tangle, :results) often breaks org structure ❌ Multiple silos (~/org/, ~/claude-memory/) cause confusion

With This Skill

✅ Denote domain knowledge bundled (file naming, frontmatter, links) ✅ Python scripts for efficient operations (finder, graph, executor) ✅ Knowledge graph cached and reusable ✅ Safe code execution (org-babel compatible) ✅ Silo management (multiple knowledge domains)

Not a file converter. An operational knowledge system.

🎯 What is Denote?

Denote is a simple yet powerful note-taking system for Emacs:

File naming convention: YYYYMMDDTHHMMSS--title__tag1_tag2.org
Knowledge graph: [[denote:TIMESTAMP]] links
Silo concept: Separate directories for different knowledge domains
Org-mode foundation: Headings, properties, timestamps, code blocks

Created by: Protesilaos Stavrou

📊 Scale: Production-Validated

This skill is built for real-world PKM systems, not toy examples:

✅ 3,000+ org files (validated at scale)
✅ Multiple silos (~/org/, ~/claude-memory/, project docs/)
✅ Denote knowledge graph (thousands of [[denote:ID]] links)
✅ Daily usage with Claude Code
✅ Part of 9-layer "Tools for Life" system

🏗️ Part of the 9-Layer System

This skill is Layer 3 (Knowledge Understanding) in a complete AI-human collaboration architecture:

Layer 7: Knowledge Publishing (notes.junghanacs.com)
Layer 6: Agent Orchestration (meta-config)
Layer 5a: Migration (memex-kb)
Layer 5b: Life Timeline (memacs-config)
Layer 4: AI Memory (claude-config)
Layer 3.5: RAG/Vector DB (embedding-config)
Layer 3: Knowledge Mgmt (Zotero, denote-org 🆕)
Layer 2: Dev Environment (Doom Emacs)
Layer 1: Infrastructure (NixOS)

Related projects:

claude-config - AI Memory System with PARA methodology
memex-kb - Universal KB Migration (Google Docs, Dooray...)
embedding-config - Vector RAG (9,477 blocks, 85% success rate)

Philosophy: Same as Anthropic Life Sciences - AI as Partner, not Tool

🚀 Real Use Cases

User: "Find all files linked to 20251021T105353"
Claude: Uses denote_graph.py
Result: Connected files in <1 second (not repeated token searches)

Case 2: Tag-based Search Across 3,000 Files

User: "All llmlog files from October 2025"
Claude: Uses denote_finder.py --tags llmlog --date 202510*
Result: Instant results with metadata

Case 3: Silo Management

User: "Is this in ~/org/ or ~/claude-memory/?"
Claude: Uses denote_silo.py find_in_silos()
Result: Resolves ambiguity automatically

Case 4: Literate Programming

User: "Execute code blocks in this org file"
Claude: Uses org_execute.py (org-babel compatible)
Result: Safe execution, structure preserved, :results updated

Case 5: Denote Link Resolution

File contains: [[denote:20251021T105353]]
Claude: Uses denote_links.py resolve_link()
Result: ~/org/llmlog/20251021T105353--full-filename.org

💡 Features

Denote Core

✅ File naming: 20251021T105353--title__tag1_tag2.org
✅ Frontmatter parsing (title, date, filetags, identifier)
✅ Link resolution ([[denote:ID]])
✅ Tag-based search
✅ Silo management (multiple knowledge domains)
✅ Knowledge graph (3,000+ files validated)

Org-mode Base

✅ Heading/property parsing
✅ Code block execution (literate programming)
✅ Timestamp handling
✅ Export (markdown, PDF, HTML)
✅ Table operations

Performance

✅ Python scripts (not token generation)
✅ Caching for repeated operations
✅ Efficient graph traversal
✅ Batch processing support

📦 Installation

Prerequisites

# Python 3.8+
pip install orgparse pypandoc pyyaml

# Pandoc (for export)
# Install via your package manager

Claude Code

git clone https://github.com/junghan0611/orgmode-skills.git
cd orgmode-skills
/plugin add .

Claude API

# See API documentation for skills integration
import anthropic

client = anthropic.Client(api_key="your-api-key")
# Skills API usage

🔧 Quick Start

Example 1: Find Denote File

from scripts.denote_finder import find_denote_file

# By identifier
filepath = find_denote_file("20251021T105353", silos=["~/org/"])
# Returns: ~/org/llmlog/20251021T105353--full-filename.org

# By tags
files = find_by_tags(["llmlog", "denote"], silo="~/org/")

Example 2: Extract Knowledge Graph

from scripts.denote_graph import build_knowledge_graph, get_connected_nodes

# Build graph (cached)
graph = build_knowledge_graph("~/org/")

# Get connected files
connected = get_connected_nodes("20251021T105353", hops=2)

Example 3: Execute Org Code Blocks

from scripts.org_execute import execute_org_blocks

# Execute all code blocks (respects :tangle, :results)
results = execute_org_blocks("file.org")

📂 Project Structure

orgmode-skills/
├── README.md                    # This file
├── README-KO.md                 # Korean version
├── LICENSE                      # Apache 2.0
├── SKILL.md                     # Claude's main guide
├── docs/                        # Development logs (Denote format)
│   └── 20251021T113500--*.org
├── denote-core.md               # Denote specification
├── denote-silo.md               # Silo management
├── denote-knowledge-graph.md    # Graph operations
├── orgmode-base.md              # Org-mode features
├── literate.md                  # Literate programming
├── export.md                    # Export capabilities
└── scripts/
    ├── denote_finder.py         # File search
    ├── denote_links.py          # Link resolution
    ├── denote_silo.py           # Silo management
    ├── denote_graph.py          # Knowledge graph
    ├── org_parser.py            # Org parsing
    ├── org_execute.py           # Code execution
    ├── org_export.py            # Export
    └── requirements.txt

🗺️ Roadmap to 0.1

Phase 1: Core (Current)

Project setup
README.md, README-KO.md
Development log (docs/)
SKILL.md (Denote-focused)

Phase 2: Denote Core

denote_finder.py (file search)
denote_links.py (link resolution)
denote_silo.py (silo management)
denote-core.md documentation

Phase 3: Org-mode Base

org_parser.py (orgparse)
org_execute.py (code blocks)
orgmode-base.md documentation

Phase 4: 0.1 Release

Testing with 3,000+ files
Examples
Public release

🤝 Contributing

This project is validated with real-world 3,000+ file Denote PKM systems. Denote users welcome!

📖 Philosophy: Tools for Life

From the creator:

"저에게 필요한 것은 경쟁력 있는 지식 노동자로 살 수 있는 방법이었습니다."

"도구는 존재에 녹아든다"

This skill embodies the "Tools for Life" philosophy - extending Anthropic's Life Sciences approach from biology research to personal knowledge work.