paper-figure

Generates publication-quality matplotlib figures and LaTeX tables from experiment data files.

Updated Apr 1, 2026
One-click install
npx skills add https://github.com/zniihgnexy/vid_tokenizer --skill paper-figure-zniihgnexy
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: paper-figure
Source: https://github.com/zniihgnexy/vid_tokenizer/tree/main/.claude/skills/paper-figure
Command: npx skills add https://github.com/zniihgnexy/vid_tokenizer --skill paper-figure-zniihgnexy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires matplotlib.

What problem does it solve? Turning raw experiment results into consistent, publication-ready figures and tables for academic papers is repetitive and error-prone, with mismatched fonts, rasterized output, and inconsistent styling across plots. ## Core Features & Use Cases - Data-driven figure generation: Creates line plots, bar charts, scatter plots, heatmaps, box plots, and multi-panel subfigures from JSON or CSV experiment data using a shared matplotlib style configuration. - LaTeX table and include snippets: Produces standalone comparison tables as .tex files and a latex_includes.tex file with ready-to-paste figure environments. - Quality review workflow: Runs a reviewer-model pass over figure captions and plans, plus a publication checklist covering grayscale readability, colorblind safety, and vector PDF output. - Use Case: After finishing experiments for a conference submission, point the skill at your results JSON files and it generates all reproducible figure scripts, 300 DPI vector PDFs, and LaTeX snippets while preserving manually created architecture diagrams. ## Quick Start Ask the AI to generate all paper figures from the experiment results in the figures directory following the plan in PAPER_PLAN.md.

Frequently Asked Questions about paper-figure

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate publication-quality matplotlib figures for a paper?▼

Use a shared style configuration that sets serif fonts, 300 DPI, tight bounding boxes, and PDF output, then write one reproducible script per figure that reads data from JSON or CSV files. Keep titles out of figures and put them in LaTeX captions instead.

What figure types can be auto-generated from experiment data?▼

Line plots for training curves, bar charts for method comparison, scatter plots, heatmaps, box and violin plots, and multi-panel subfigure grids can all be generated from data. Architecture diagrams and qualitative image grids must be created manually.

Should I use PDF or PNG for LaTeX paper figures?▼

PDF is the recommended format because it keeps text and lines as vector graphics that scale cleanly in print. PNG at 300 DPI is only a raster fallback when vector output is not possible.

Can this skill create model architecture diagrams?▼

No, architecture and pipeline diagrams are outside its scope and should be drawn manually with tools like draw.io, Figma, or TikZ. The skill preserves existing manual figures in the figures directory and only generates the LaTeX include snippet for them.

How do I make matplotlib figures colorblind-friendly?▼

Choose a colorblind-safe palette such as the colorblind option instead of tab10, and verify figures remain distinguishable in grayscale for print. Avoid relying on color alone by varying line styles or markers.