An Agent Skill that turns documents and code repositories into interactive structure maps.
- Documents: Analyze Markdown, LaTeX, and PDF files to map key ideas, methods, evidence, and their relationships.
- Code repositories: Explain Python project structure, file responsibilities, and how components work together.
- Interactive viewer: Explore results by expanding nodes, navigating files, searching, and switching between English and Chinese.
- GPT code structure map: Explore the structure of GPT model code.
- Attention Is All You Need paper structure map: Explore the paper's main ideas and their relationships.
Click either link to open the interactive page directly in your browser. Each example is also available as a self-contained HTML file in examples/ for offline use.
npx skills add AtheavenMa/ExtractStructureAsk your AI agent:
Use extract-structure to visualize this paper: path/to/paper.pdf
Use extract-structure to map this project: path/to/repository/
The agent reads your selected files and creates an interactive map. Open it in a local browser viewer or save it as a self-contained HTML page to share.
Python 3.10+ and a browser. The included graph tools use Python's standard library; PDF reading depends on the host agent's available tools.
Copyright (c) 2026 Haotian Ma.
This project is licensed under the MIT License. See LICENSE for details.
If you use ExtractStructure in your work, please cite the Zenodo release:
@software{ma2026extractstructure,
author = {Ma, Haotian},
title = {ExtractStructure: An Agent Skill for Interactive Structure Visualization},
year = {2026},
version = {0.1.0},
publisher = {Zenodo},
doi = {10.5281/zenodo.24051078},
url = {https://doi.org/10.5281/zenodo.24051078}
}