This skill is specifically designed to integrate Codex's native drawing capabilities with the Gemini image generation and PubMed academic figure retrieval provided by MCP.
Core Strategies
-
Hybrid Rendering
- Structured and Flowchart: Prefer using Codex's native Markdown mermaid syntax, SVG, or Python matplotlib/plotly to draw precise architecture diagrams, bar charts, scatter plots, and data models.
- Complex Medical/Biological Illustrations: Use MCP tools (such as mcp_academic-figu_generate_figure or Gemini tools) to generate highly complex 3D anatomical diagrams, histological images, or irregular concept diagrams.
- Dual Validation: If needed, use mcp_academic-figu_plan_figure to plan the most appropriate rendering route (deciding whether programmatic drawing or AI generation should be used).
-
Citation-Ready & Provenance
- Cited sources must be accurate; PubMed tools unified_search or get_article_figures can be used to request reference sources or example images.
- Generated text captions must comply with academic journal standards (including figure number, short title, detailed description, and PMID source).
-
Iterative Refinement
- After generating a preliminary figure, use the interactive tools mcp_academic-figu_evaluate_figure (8-dimension quality assessment) or mcp_academic-figu_edit_figure to make fine adjustments based on user feedback (e.g., changing colors, adjusting layout).
- For native SVG/Mermaid code, directly use Codex's text edit capability and replace_string_in_file to refactor and update.
Workflow
- Step 1: Plan the Figure (Plan) After obtaining the requirements or literature content, analyze the appropriate figure type (figure_type: flowchart, mechanism, comparison, data_visualization, etc.).
- Step 2: Route Selection
- If it can be expressed programmatically (such as statistical data), guide Codex to write a Python script or Mermaid.
- If a polished raster image is needed, call generate_figure.
- Step 3: Add Context and Assemble
Generate the Markdown embed format for the final image
, and add complete academic sources and PICO/MeSH background explanation.