Figures Diagram

Codex skill for academic non-numeric diagram generation and OpenRouter-backed rendering.

This repository packages a reusable workflow for turning diagram ideas into polished English prompts for image models and, when requested, directly rendering publication-oriented draft figures.

At a Glance

• Prompt engineering for flowcharts, frameworks, concept maps, taxonomies, and timelines
• Academic defaults for layout, readability, color usage, and labeling
• Prompt-only mode for reusable figure ideation
• Production mode for direct OpenRouter image rendering
• Environment-variable based credential handling instead of hardcoded secrets

Preview

Architecture

flowchart LR
    A["User Diagram Request"] --> B["Figure Type Classifier"]
    B --> C["Input Inference and Defaults"]
    C --> D["Prompt Construction Engine"]
    D --> E["Prompt-Only Output"]
    D --> F["OpenRouter Production Renderer"]
    F --> G["Image File / Prompt File / JSON Response"]

Loading

Best Fit

Use this skill for:

• flowcharts
• process diagrams
• architecture diagrams
• model structure diagrams
• concept maps
• theoretical framework diagrams
• organization charts
• taxonomy diagrams
• timelines

Do not use it for numeric charts such as line charts, bar charts, scatterplots, or heatmaps.

Workflow

1. Classify the figure type
2. Gather or infer the minimum inputs
3. Choose prompt-only or production mode
4. Build a polished academic prompt
5. Render through OpenRouter when direct generation is requested

Repository Layout

SKILL.md
agents/
  openai.yaml
references/
  openrouter-production.md
  prompt-templates.md
  style-and-tools.md
scripts/
  render_via_openrouter.py
assets/
  banner.svg
  preview.svg

Quick Start

Install into your Codex skills directory:

$CODEX_HOME/skills/figures-diagram

Typical Windows path:

C:\Users\<your-user>\.codex\skills\figures-diagram

Example prompts:

• Use $figures-diagram to create an academic concept map prompt for my theoretical framework.
• Use $figures-diagram to generate a methodology flowchart prompt.
• Use $figures-diagram to render a research-process figure now.

Production Rendering

The bundled script renders prompts through OpenRouter and saves:

• the generated image
• the original prompt text
• the raw JSON response
• any assistant text returned by the model

Example:

python scripts/render_via_openrouter.py --prompt "Draw a publication-ready academic flowchart illustrating the proposed research process from data collection to model evaluation." --aspect-ratio 4:3 --output-dir output/figures

Default production model:

• google/gemini-3.1-flash-image-preview

User-facing aliases may refer to this as:

• Nano Banana
• Nano Banana 2

Environment Variables

• OPENROUTER_API_KEY
• OPENROUTER_IMAGE_MODEL

The rendering script also supports Windows user environment variables.

Companion Repository

• Thesis-Pipeline for proposal drafting, literature retrieval, and methodology planning

Design Principles

• Keep labels crisp and readable
• Prefer uncluttered academic composition over decorative effects
• Use explicit layout and relationship instructions
• Keep prompts drawable and production-oriented
• Use prompt-only mode when direct rendering is unnecessary

License

MIT