baikemark/happy-figure-skill

★ 13PythonAudience · researcherComplexity · 2/5Setup · moderate

Mindmap

mindmap
  root((repo))
    What It Does
      Scientific figure prompts
      Research content input
      Model-specific output
    Research Domains
      ML architecture diagrams
      Biomedical illustrations
      Earth science visuals
    How It Works
      Agent skill invocation
      Prompt routing
      Figure structure compile
    Compatible Models
      GPT Image 2
      Qwen Image 2
      Five model comparison
    Audience
      Academic researchers
      Datawhale community

mindmap root((repo)) What It Does Scientific figure prompts Research content input Model-specific output Research Domains ML architecture diagrams Biomedical illustrations Earth science visuals How It Works Agent skill invocation Prompt routing Figure structure compile Compatible Models GPT Image 2 Qwen Image 2 Five model comparison Audience Academic researchers Datawhale community

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Describe a machine learning architecture or training pipeline and get a detailed, model-specific prompt ready to paste into an image generator.

USE CASE 2

Generate a biomedical process illustration or chemistry mechanism diagram from a figure caption or study plan.

USE CASE 3

Compare how the same scientific figure prompt renders across five different image generation models to choose the best visual output.

Tech stack

Python

Getting it running

Difficulty · moderate Time to first run · 30min

Runs as an AI agent skill invoked by Claude Code or a compatible assistant, an image generation API is needed separately to produce the actual images.

In plain English

Happy Figure Skill is a plugin for AI coding assistants like Claude Code that helps researchers create scientific illustrations using AI image generation models. The README is primarily in Chinese. Instead of generating images directly, this tool takes research content such as a paper, figure caption, or study plan and produces a structured text prompt that you can then feed into an image generation model like GPT Image 2, Qwen Image 2, or similar tools. The core idea is that writing a good prompt for a scientific diagram requires knowing the domain, the type of figure you want, and the particular quirks of the image model you are using. Happy Figure Skill handles that translation step. You describe what your research is about and what kind of figure you need, and the skill produces a detailed, model-specific prompt rather than a generic one. The README describes it as a prompt router and figure structure compiler rather than a one-size-fits-all prompt. The tool covers several research domains including computer science and machine learning architecture diagrams, materials chemistry mechanism diagrams, biomedical process illustrations, and earth science visualizations. The showcase section of the README shows the same five tasks run through five different image models side by side, demonstrating how the same structured prompt produces different visual styles on different models, and how factors like text rendering stability and diagram layout vary across them. Happy Figure Skill is built as an agent skill, meaning it is not a standalone program you run directly. It is designed to be invoked by an AI agent such as Claude Code, which reads your research content and calls the skill to generate the prompt. The skill is part of a broader project called Happy Figure maintained by the Datawhale community. The README links to a QQ group for Chinese-speaking users who want to discuss or get help with the tool.

Copy-paste prompts

Prompt 1

I have a paper about a transformer-based anomaly detection model and I need an architecture diagram. Invoke happy-figure-skill with my figure caption and generate a prompt optimized for GPT Image 2.

Prompt 2

I need a biomedical process illustration showing how a virus enters a cell. Describe my research context to happy-figure-skill and produce a structured prompt I can feed into Qwen Image 2.

Prompt 3

I want to see how the same materials chemistry mechanism diagram looks across GPT Image 2, Qwen Image 2, and three other models. Use happy-figure-skill to generate five model-specific prompts from my figure description so I can compare outputs.

Open on GitHub → Explain another repo

← baikemark on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.