Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
OpenTikZ turns a verbal description of a diagram into a working TikZ file by matching the request against an indexed library of neural net architectures, pipelines, system blocks, and icons. The skill is designed for researchers drafting conference figures, LaTeX authors who edit diagrams programmatically, and AI agents that need to produce structured visuals. It enforces standalone compilation, palette-based colors, and a copy-and-edit workflow that keeps the library pristine while the user gets a figure that compiles.
WHO IT'S FOR
ML/NLP researchers
produce publication-ready TikZ diagrams
LaTeX / technical authors
create and edit TikZ figures programmatically
AI agent / automation builders
integrate structured diagram generation into workflows
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Installs: Complete cloned OpenTikZ repository, including the using-opentikz skill and the repository's icons, templates, and examples. · Codex, Cursor, Gemini CLI
Before you start
Git
An AI agent that can read local repository files
Clone this repository into a local directory using Git. The evidence names the operation but does not provide a complete terminal command, so no command is supplied here.
Tell Codex, Cursor, or Gemini CLI to use skills/using-opentikz/SKILL.md from the cloned repository. Work from your own paper or LaTeX project's directory so generated output is placed there rather than in the OpenTikZ library.
The actual LaTeX project directory to substitute for <PROJECT_PATH>
Any desired labels, colors, dimensions, or publication-column constraints; otherwise the existing template styling can be retained
Use the using-opentikz skill in my LaTeX project at <PROJECT_PATH>. Find the OpenTikZ encoder-decoder template, copy it into my project, add one hidden layer, and verify that the edited copy compiles. Treat <PROJECT_PATH> as a placeholder for my actual project directory, and ask me about any material ambiguity before editing.