drchronx/ai-agent-research-starter-kit/ai4scholar-sci-draw

AI4Scholar Sci Draw

Use this skill for scientific figure generation or figure refinement through AI4Scholar sci_draw.

Core Rules

• Do not invent experimental results, data values, brain maps, statistical annotations, or electrode effects.
• For data figures, use the user's real data and deterministic plotting tools first; use sci_draw for schematics, conceptual figures, mechanism diagrams, or style refinement.
• Keep figure text short and publication-readable.
• Preserve scientific meaning during style transfer or editing.

Tool From The Source Guide

Good Use Cases

• Graphical abstract for a manuscript.
• EEG/ERP experiment procedure diagram.
• BCI pipeline figure: acquisition, preprocessing, feature extraction, classifier, feedback.
• Management or psychology scenario-experiment flow.
• Text-mining or machine-learning architecture schematic.
• Polishing an existing conceptual figure into journal style.
• Converting a rough sketch into SVG-like vector style.

Figure Workflow

1. Identify figure type:

• mechanism,
• workflow,
• architecture,
• experimental procedure,
• graphical abstract,
• review framework,
• visual polishing.

1. Extract required scientific content:

• constructs,
• variables,
• sample/stimuli,
• pipeline steps,
• model components,
• labels.

1. Draft a compact prompt.
2. Generate or edit using sci_draw.
3. Review:

• scientific accuracy,
• label correctness,
• readability,
• color accessibility,
• consistency with target journal.

1. Iterate with specific correction prompts.

Prompt Template

Use AI4Scholar sci_draw to generate a publication-style scientific figure.
Figure type: [mechanism diagram / workflow / BCI pipeline / EEG experiment procedure / graphical abstract]
Research topic:
Required elements:
Forbidden elements:
Language: [Chinese / English]
Style: [minimal journal style / neuroscience style / management theory model / machine-learning architecture]
Output requirement: [PNG / SVG / editable vector preferred]

Figure Review Checklist

Common Corrections

Keep the original structure unchanged, translate labels into English, and reduce decorative elements.

Make EEG preprocessing, feature extraction, classifier, and feedback a horizontal workflow. Do not add nonexistent data results.

Convert the figure into a theory-mechanism model suitable for an SSCI management paper. Preserve variable names and arrow directions.