This project addresses the common scientific research pain point of selecting the appropriate chart type to effectively communicate data-driven conclusions, rather than just the technical execution of plotting. It acts as an intelligent advisor and renderer for publication-grade scientific figures, targeting researchers and power users. The primary benefit is ensuring visualizations are not only aesthetically correct but also statistically sound and aligned with journal standards, reducing submission friction.

How It Works

The core approach is "think first, plot second," deviating from generic plotting tools. It employs an 8-step workflow: understanding the argument, profiling data (types, distributions, correlations), selecting the optimal chart based on data characteristics and intended message (with active interception of suboptimal choices), adhering to journal specifications, configuring styles (including CJK fonts), plotting, performing a rigorous self-check, and exporting. This iterative self-check loop, new in v2.1, combines programmatic audits for deterministic issues like clipping and overlap with AI-driven perceptual checks for elements like legend occlusion and panel alignment.

Quick Start & Requirements

• Installation: Clone the repository (git clone https://github.com/Haojae/scipilot-figure-skill.git ~/.claude/skills/scipilot-figure-skill) and install dependencies (pip install -r ~/.claude/skills/scipilot-figure-skill/requirements.txt).
• Prerequisites: Python 3.9+ recommended.
• Optional Dependencies: SciencePlots, pypdf, kaleido, PyMuPDF for enhanced features.
• Links: GitHub Repository

Highlighted Details

• Intelligent Chart Selection: Actively guides users away from inappropriate plots (e.g., mean bars for n<10, pie charts) towards better alternatives like stripplots or box plots.
• Publication-Grade Output: Enforces journal-specific standards for size, font, DPI, and colorblind-safe palettes, rendering figures at their final intended dimensions.
• Visual Self-Check Loop (v2.1): Post-rendering audits catch issues like missing glyphs, clipped text, overlapping labels (programmatic) and legend-data overlap, panel alignment, and grayscale distinctness (AI-driven).
• CJK Font Support: Automatic configuration for Chinese characters and handling of minus signs to prevent rendering issues in non-English contexts.
• Hard Rules: Prioritizes vector formats, readable type sizes, clear error reporting in captions, and avoids common pitfalls like dual-Y axes for correlation.

Maintenance & Community

The project is licensed under MIT and dated 2026, authored by Haojae. Other related SciPilot skills are listed as "Planned," indicating ongoing development within a broader ecosystem. No specific community channels (e.g., Discord, Slack) or detailed roadmap beyond the planned skills are provided in the README.

Licensing & Compatibility

The project is released under the MIT License, which permits commercial use, modification, distribution, and patent use, with minimal restrictions beyond attribution.

Limitations & Caveats

The README focuses on the capabilities and workflow, with no explicit mention of alpha status, known bugs, or unsupported platforms. Its integration as a "Skill" for Claude Code suggests its primary use case might be within that specific AI assistant environment.