NoteMD Pro - Concept Extraction (Agentic Workflow)

Overview

This skill transforms standard Markdown text into an interconnected Knowledge Graph by exhaustively extracting every relevant concept and generating atomic notes with proper backlinks.

⚠️ Critical Rule: Exhaustive Extraction

When executing this skill, DO NOT STOP at 3 to 5 concepts. You must thoroughly scan the entire document and extract every single core knowledge point, technical term, scientific principle, proper noun, and specialized vocabulary word. A comprehensive document (e.g., 1000 words) might yield 15-30 distinct concepts.

Your goal is to be as complete and reasonable as possible, ensuring the end-user's Knowledge Graph is densely populated and highly interconnected.

Step-by-Step Agent Instructions

When the user asks you to extract concepts from a specific Markdown file (or text), follow these steps strictly:

Step 1: Comprehensive Parsing

Read the entire target document thoroughly. Mentally (or in your scratchpad) list every concept following these criteria:

• Scientific & Technical Terms: e.g., "DNA Replication", "Ribosome", "Stellar Nucleosynthesis", "Isotope", "Spectroscopy", "Amino Acid".
• Theories & Principles: e.g., "Statistical Mechanics", "Conservation of Energy".
• Proper Nouns & Tools: e.g., "Tesseract OCR", "Python multiprocessing", "Markdown".
• Abstract Concepts: Limit abstract ideas unless they are central to the domain. Focus on specific noun-phrases.
• Normalization: Always normalize to the singular form (e.g., "Isotope" instead of "Isotopes") and capitalize Title Case (e.g., "DNA Replication").

Step 2: Generate Atomic Concept Notes

For every single concept you identified in Step 1, create a separate atomic note file in the target directory (often a Concepts/ folder or the same folder as the source).

Filename: [Concept Name].md

Content Format:

# [Concept Name]

Brief 1-2 sentence definition or summary of the concept (derive this from your internal knowledge base or the source text context).

## Linked From

- [[Name of the Original Source File]]

(Note: Always append the exact name of the file you are extracting from to the ## Linked From section, wrapped in [[wiki-links]])

Step 3: Inject Wiki-Links into Source

After creating the atomic concept notes, you must run the link-analyzer skill (or do it manually if instructed) to go back into the original source file and wrap the occurrences of those concepts with Obsidian wiki-links: [[Concept Name|original text]].

Output Quality Checklist for the Agent

• Did I extract all possible valid concepts, or did I stop early out of laziness? (Aim for exhaustive depth).
• Are the concepts normalized to singular, Title Case forms?
• Did I create an individual .md file for every extracted concept?
• Did I add the ## Linked From backlink in every concept note?