Import Obsidian Vault (Justice Naturelle)

This skill imports propositions and relations from an Obsidian vault into the simple-kb knowledge base with multilingual support (French, English, Spanish).

Requirements

• •Node.js with npm installed
• •@anthropic-ai/sdk: Install with npm install @anthropic-ai/sdk
• •ANTHROPIC_API_KEY: Set in .env file for translations
• •OBSIDIAN_VAULT_IMPORT_FOLDER: Set in .env file

Instructions

You are tasked with importing content from an Obsidian vault into the simple-kb format.

Input Files

The skill reads the base folder path from the .env file:

• •Environment Variable: OBSIDIAN_VAULT_IMPORT_FOLDER (required)

The skill imports all .md files from:

• •{OBSIDIAN_VAULT_IMPORT_FOLDER}/Propositions/ (all subdirectories)
• •{OBSIDIAN_VAULT_IMPORT_FOLDER}/Définitions/
• •{OBSIDIAN_VAULT_IMPORT_FOLDER}/Citations/
• •{OBSIDIAN_VAULT_IMPORT_FOLDER}/*.md (root level files)

Understanding the Structure

The Obsidian vault contains markdown files with:

• •Optional YAML frontmatter with tags
• •Main content in markdown format
• •Optional relation tables at the end

Relation tables format:

| Relation     | Proposition               | Détail                        |
| ------------ | ------------------------- | ----------------------------- |
| Implication  | [[Target File Name]]      | Description text              |
| Illustration | [[Another Target]]        | More description              |

Extraction Process

1. •

   Build UUID map (CRITICAL STEP):

   • •Scan all .md files in the vault
   • •Generate ONE UUID for each file
   • •Store in a map: { filename_without_extension → uuid }
   • •This ensures all three languages use the same UUID for the same proposition
2. •

   Parse all files:

   • •Extract title from filename (remove .md extension)
   • •Determine proposition type based on folder and tags
   • •Extract content (remove YAML frontmatter and relation tables)
   • •Parse relation tables to extract relations
3. •

   Create proposition files in /propositions/{uuid}/:

   code

   /propositions/
     a1b2c3d4/
       fr-title-slug.md
       en-title-slug.md
       es-title-slug.md
   

   Each file contains:

   yaml

   ---
   title: "Original Title"
   type: [définition|performative|prescriptive|descriptive|citation] (or omit if no type)
   language: [fr|en|es]
   tags: [preserved from source YAML]
   created: [current date]
   modified: [current date]
   author: TBD
   status: imported
   source_file: "original_filename.md"
   source_folder: "folder/path"
   ---
   
   [CONTENT in respective language]
   

   Language handling:

   • •French: Original content from source file
   • •English: Translated using Claude AI
   • •Spanish: Translated using Claude AI
4. •

   Create relation files in /relations/{source-uuid}-{target-uuid}/:

   code

   /relations/
     a1b2c3d4-e5f6g7h8/
       fr-a1b2c3d4-e5f6g7h8.md
       en-a1b2c3d4-e5f6g7h8.md
       es-a1b2c3d4-e5f6g7h8.md
   

   • •For each relation in the table, create a relation file
   • •Resolve WikiLinks [[Target]] to target file's UUID
   • •Skip relations where target file doesn't exist
   • •Include "Détail" column text in relation content

Type Determination

Proposition types are determined by:

• •Files in Définitions/ folder → type: "définition"
• •Files in Citations/ folder → type: "citation"
• •Files with YAML tag tags: [performatif] → type: "performative"
• •Files with YAML tag tags: [définition] → type: "définition"
• •Other files → omit type field entirely

Slug Generation

Slugs are generated from filenames:

• •Remove .md extension
• •Convert to lowercase
• •Remove accents (é→e, è→e, à→a, etc.)
• •Replace spaces and special characters with hyphens
• •Keep numbers
• •Maximum 50 characters
• •Examples:
  • •"1. Vie et intégrité physique.md" → "1-vie-et-integrite-physique"
  • •"Condamnation de l'esclavage.md" → "condamnation-de-l-esclavage"

Content Extraction

From each source file:

1. •Remove YAML frontmatter (everything between --- markers at the start)
2. •Remove relation tables (tables with columns: Relation, Proposition, Détail)
3. •Remove horizontal rules (---) that separate sections
4. •Keep all other markdown content including URLs

Example transformation:

---
tags:
  - performatif
---
Main content here.

Some more text.
https://example.com/link

---

| Relation | Proposition | Détail |
| -------- | ----------- | ------ |
| Implication | [[Target]] | Description |

Becomes:

Main content here.

Some more text.
https://example.com/link

Important Notes

• •UUID consistency: CRITICAL - The same source file must use the same UUID across all languages
• •Preserve tags: Keep original YAML tags from source files in the generated propositions
• •Translation: Use Claude AI to translate French content to English and Spanish
• •Validate targets: Skip relations where the target WikiLink doesn't resolve to an existing file
• •No reverse relations: Only create relations as explicitly declared in the tables

Output

After import, report:

• •Number of unique propositions
• •Number of proposition files created per language (fr, en, es)
• •Number of relation files created per language (fr, en, es)
• •Any errors or warnings encountered (missing targets, parsing issues, etc.)

Future Improvements (Deferred)

• •Support for updating existing propositions
• •Validation of relation types
• •Detection of circular dependencies
• •Support for additional metadata fields