obsidian-llm-wiki: Skill for Claude Code

Tutor Setup — Knowledge to Obsidian StudyVault

CWD Boundary Rule (ALL MODES)

NEVER access files outside the current working directory (CWD). All source scanning, reading, and vault output MUST stay within CWD and its subdirectories. If the user provides an external path, ask them to copy the files into CWD first.

Mode Detection

On invocation, detect mode automatically:

1. Check for project markers in CWD:
   • package.json, pom.xml, build.gradle, Cargo.toml, go.mod, Makefile, *.sln, pyproject.toml, setup.py, Gemfile
2. If any marker found → Codebase Mode
3. If no marker found → Document Mode
4. Tie-break: If .git/ is the sole indicator and no source code files (*.ts, *.py, *.java, *.go, *.rs, etc.) exist, default to Document Mode.
5. Announce detected mode and ask user to confirm or override.

---

Document Mode

Transforms knowledge sources (PDF, text, web, epub) into study notes. Templates: templates.md

Phase D1: Source Discovery & Extraction

1. Auto-scan CWD for **/*.pdf, **/*.txt, **/*.md, **/*.html, **/*.epub (exclude node_modules/, .git/, dist/, build/, StudyVault/). Present for user confirmation.
2. Extract text (MANDATORY tools):
   • PDF → pdftotext CLI ONLY (run via Bash tool). NEVER use the Read tool directly on PDF files — it renders pages as images and wastes 10-50x more tokens. Convert to .txt first, then Read the .txt file.

     pdftotext "source.pdf" "/tmp/source.txt"
     

   • If pdftotext is not installed, install it first: brew install poppler (macOS) or apt-get install poppler-utils (Linux).
   • URL → WebFetch
   • Other formats (.md, .txt, .html) → Read directly.
3. Read extracted .txt files — understand scope, structure, depth. Work exclusively from the converted text, never from the raw PDF.
4. Source Content Mapping (MANDATORY for multi-file sources):
   • Read cover page + TOC + 3+ sample pages from middle/end for EVERY source file
   • NEVER assume content from filename — file numbering often ≠ chapter numbering
   • Build verified mapping: { source_file → actual_topics → page_ranges }
   • Flag non-academic files and missing sources
   • Present mapping to user for verification before proceeding