# Difference Between DEPTH=reference and DEPTH=study in book-to-skill > Understand the DEPTH=reference vs DEPTH=study difference in book-to-skill. Reference offers quick metadata, Study provides a full guide with every detail. Optimize your learning. - Repository: [Virgilio Junior/book-to-skill](https://github.com/virgiliojr94/book-to-skill) - Tags: deep-dive - Published: 2026-08-28 --- **DEPTH=reference generates a lightweight, fast-lookup index containing only chapter hierarchies and metadata, whereas DEPTH=study produces a comprehensive study guide embedding every paragraph, example, and annotation from the source material.** The `book-to-skill` repository provides a specialized extraction pipeline that converts books into structured skill files for AI agents. Understanding the `DEPTH` configuration flag is essential for controlling the granularity of generated output, as this parameter directly determines whether you receive a concise reference catalog or an exhaustive knowledge base suitable for deep research. ## What the DEPTH Flag Controls According to the source code in [`scripts/extract.py`](https://github.com/virgiliojr94/book-to-skill/blob/main/scripts/extract.py), the `DEPTH` environment variable acts as a content-filtering directive during the extraction phase. It instructs the parser how deeply to traverse the document tree before serializing content into the final [`SKILL.md`](https://github.com/virgiliojr94/book-to-skill/blob/main/SKILL.md) artifact. The flag supports two mutually exclusive modes that trade storage efficiency against informational completeness. ## DEPTH=reference: Lean Index Mode When you set `DEPTH=reference`, the extractor preserves only the **top-level document hierarchy**—specifically chapter titles, section headers, and essential metadata. As noted in [`SKILL.md`](https://github.com/virgiliojr94/book-to-skill/blob/main/SKILL.md) at line 292, this mode creates a "lean, fast-lookup" structure that omits body text, code examples, and detailed annotations, resulting in a compact representation optimized for navigation. Use this mode when you need: - A lightweight table of contents for quick browsing and cataloging - Fast-loading skill files for memory-constrained environments - Simple reference lookups without full-text retrieval requirements ```bash DEPTH=reference python -m scripts.extract ``` ## DEPTH=study: Comprehensive Study Mode Setting `DEPTH=study` engages the full extraction pipeline, retaining **every paragraph, code example, and detailed annotation** found in the source document. The documentation at line 404 of [`SKILL.md`](https://github.com/virgiliojr94/book-to-skill/blob/main/SKILL.md) demonstrates that this mode generates a verbose skill file capable of supporting deep-dive interactions, contextual question answering, and AI tutoring applications where complete context is required. Use this mode when you require: - Exhaustive content coverage for research-grade retrieval systems - In-context learning capabilities for detailed tutoring agents - Complete preservation of examples and explanatory text ```bash DEPTH=study python -m scripts.extract ``` ## Programmatic Configuration Options Beyond shell environment variables, you can configure the depth programmatically within Python applications. The `book_to_skill` package reads the `DEPTH` value during initialization via [`book_to_skill/__init__.py`](https://github.com/virgiliojr94/book-to-skill/blob/main/book_to_skill/__init__.py), allowing dynamic configuration before invoking the extraction pipeline. ```python import os from book_to_skill import extractor # Configure depth programmatically os.environ["DEPTH"] = "study" # or "reference" skill = extractor.run("") # Output contains either lean reference or full study content print(skill.metadata.depth) # Confirms active extraction mode ``` ## Implementation Details and Source Files The semantic distinction between these modes is formally documented in [`SKILL.md`](https://github.com/virgiliojr94/book-to-skill/blob/main/SKILL.md), which explicitly defines the two options and provides a comparison table at line 347 outlining their respective use cases. The extraction logic that implements these behaviors resides in [`scripts/extract.py`](https://github.com/virgiliojr94/book-to-skill/blob/main/scripts/extract.py), which inspects the environment variable and adjusts its AST traversal depth before the generator serializes the final artifact. ## Summary - **DEPTH=reference** produces compact, hierarchical indexes containing only chapter and section metadata, ideal for quick navigation and catalog applications - **DEPTH=study** generates comprehensive skill files retaining every paragraph, example, and annotation, enabling deep research and tutoring use cases - Configure the flag via environment variable, command line, or programmatically through the Python API before calling the extractor - The parameter is processed in [`scripts/extract.py`](https://github.com/virgiliojr94/book-to-skill/blob/main/scripts/extract.py) and documented in [`SKILL.md`](https://github.com/virgiliojr94/book-to-skill/blob/main/SKILL.md) with specific line references defining each mode's behavior ## Frequently Asked Questions ### Which DEPTH mode produces smaller output files? **DEPTH=reference** generates significantly smaller files because it deliberately discards body text and retains only the structural metadata necessary for cataloging and navigation. ### Can I convert an existing reference skill file to a study skill file? No, you must reprocess the original book source with `DEPTH=study` specified to capture the full text content that was omitted during the initial reference extraction. ### Does DEPTH=study impact processing performance? Yes, `DEPTH=study` requires substantially more processing time and memory because it traverses the entire document tree and serializes significantly more content into the final skill structure compared to the lean reference mode. ### Where is the DEPTH parameter documented in the repository? The parameter semantics are documented in [`SKILL.md`](https://github.com/virgiliojr94/book-to-skill/blob/main/SKILL.md) at line 292, with a detailed comparison table listing both modes side-by-side at line 347 and worked examples demonstrating the output differences at line 404.