python-development


Custom AI agents and skills for Research Software Engineering (RSE) and Scientific Computing tasks, designed for use with Claude Code and compatible AI coding assistants.

GitHub Codespaces

Click the badge above to open a Codespace with all RSE plugins pre-installed. Claude Code and GitHub Copilot CLI are ready to use immediately.

Using a LiteLLM Gateway (Optional)

To route Copilot through a custom LiteLLM-compatible gateway, you need to set two secrets. Because GitHub Codespaces secrets can only be scoped to repos you own, follow these steps:

1. Fork this repo — Click Fork on github.com/uw-ssec/rse-plugins.
2. Add secrets — Go to your GitHub account settings: Settings → Codespaces → Secrets, then add:

- LITELLM_BASE_URL — your gateway base URL - LITELLM_API_KEY — your gateway API key - When adding each secret, under Repository access, select your fork.

1. Open a Codespace from your fork — Go to your fork on GitHub and click Code → Codespaces → Create codespace on main.

The Codespace will automatically detect the secrets and configure Copilot to route through your gateway.

Purpose

This repository provides specialized agents and skills that understand the unique challenges of scientific software development, including:

• •Modern Scientific Python development following community best practices
• •Reproducible environment management with pixi
• •Python packaging and distribution with pyproject.toml
• •Comprehensive testing strategies with pytest
• •Scientific computing workflows and numerical methods
• •Research software engineering practices
• •Domain-specific scientific computing (astronomy, geospatial analysis, climate science)
• •Interactive data visualization with the HoloViz ecosystem (Panel, hvPlot, HoloViews, Datashader, GeoViews, Lumen)
• •Scientific Python ecosystem (NumPy, Pandas, SciPy, Matplotlib, Xarray, Astropy, etc.)

Installation

To use these agents and skills in Claude Code, add this repository to your plugin marketplace:

/plugin marketplace add uw-ssec/rse-plugins

Once installed, the agents and skills will be available in your Claude Code environment and can be invoked when working on scientific software projects.

Available Plugins

The repository provides Claude Code plugins organized by domain. Each plugin contains agents (specialized AI personas) and skills (reusable knowledge modules).

Scientific Python Development Plugin

Expert agents and comprehensive skills for modern Scientific Python development.

Agents:

• •Scientific Python Expert - Comprehensive agent for scientific Python development following Scientific Python Development Guide best practices
• •Scientific Documentation Architect - Expert in creating comprehensive, user-friendly documentation for scientific software following Scientific Python community standards

Skills:

• •pixi-package-manager - Fast, reproducible scientific Python environments with unified conda and PyPI management
• •python-packaging - Modern packaging with pyproject.toml, src layout, and Hatchling build backend
• •python-testing - Robust testing strategies with pytest following Scientific Python community guidelines
• •code-quality-tools - Linting, formatting, and type checking tools for Python code quality
• •scientific-documentation - Documentation best practices for scientific software including Sphinx, API docs, tutorials, and examples

When to use: Scientific computing projects, data analysis pipelines, research software development, package creation, reproducible research workflows

Scientific Domain Applications Plugin

Domain-specific scientific computing agents and skills for astronomy, geospatial analysis, climate science, and interactive visualization.

Agents:

• •Astronomy & Astrophysics Expert - Expert in astronomical data analysis, FITS files, coordinate systems, and photometry/spectroscopy pipelines with Astropy

Skills:

• •xarray-for-multidimensional-data - Work with labeled multidimensional arrays and NetCDF/Zarr datasets for climate and Earth science
• •astropy-fundamentals - Astronomical data formats, coordinate transformations, physical units, and time handling with Astropy

When to use: Astronomy research, telescope data processing, climate data analysis, Earth science workflows, geospatial analysis

AI Research Workflows Plugin

Structured AI-enabled workflow for complex software development tasks with explicit phases for research, planning, experimentation, implementation, and validation.

Agent:

• •Research Workflow Orchestrator - Guides users through structured development workflows from research to validated implementation

Commands:

• •/research - Document and understand existing code, patterns, and architecture
• •/plan - Create detailed, testable implementation plans through interactive research
• •/iterate-plan - Refine existing plans based on feedback or changed requirements
• •/experiment - Try multiple approaches before committing to implementation (optional)
• •/implement - Execute the plan phase by phase with verification checkpoints
• •/validate - Systematically verify implementation against plan criteria

Skill:

• •research-workflow-management - Systematic workflow methodology creating auditable trail of technical decisions in .agents/ directory

When to use: Complex feature development, architectural changes, exploratory implementation, technical research tasks, systematic code refactoring, documented decision-making

Project Management Plugin

Project lifecycle management — onboarding, documentation quality, handoff readiness, and community health for research software projects in any language.

Agents:

• •Project Onboarding Specialist - Expert in project initialization, contributor onboarding, and knowledge transfer
• •Documentation Validator - Expert in documentation quality assurance, setup instruction validation, and completeness checking

Commands:

• •/setup-project - Scaffold a new project with community health files and standard structure
• •/project-handoff - Assess project readiness for handoff to new maintainers
• •/validate-project-handoff - Test that setup instructions and documentation actually work

Skills:

• •community-health-files - Templates for README, CONTRIBUTING, LICENSE, CODE_OF_CONDUCT, SECURITY, and CITATION.cff
• •documentation-validation - Validation tools (Vale, markdownlint, HTMLProofer) and documentation quality metrics

When to use: Project initialization, onboarding documentation, project handoff, documentation quality auditing, community health file creation

HoloViz Visualization Plugin

Expert agents and comprehensive skills for interactive data visualization using the HoloViz ecosystem (Panel, hvPlot, HoloViews, Datashader, GeoViews, Lumen).

Agents:

• •Panel Specialist - Expert in building interactive dashboards, web applications, and component systems with Panel and Param
• •Visualization Designer - Strategic guide for multi-library visualization design using HoloViz ecosystem tools
• •Data Engineer - Specialist in large-scale data rendering and performance optimization with Datashader (100M+ points)
• •Geo-Spatial Expert - Expert in geographic and mapping visualizations with GeoViews and spatial data handling

Skills:

• •panel-dashboards - Interactive dashboard and application development with Panel and Param
• •plotting-fundamentals - Quick plotting and interactive visualization with hvPlot
• •data-visualization - Advanced declarative visualization with HoloViews
• •advanced-rendering - High-performance rendering for large datasets with Datashader
• •geospatial-visualization - Geographic and mapping visualizations with GeoViews
• •colormaps-styling - Color management and visual styling with Colorcet
• •parameterization - Declarative parameter systems with Param for type-safe configuration
• •lumen-dashboards - Declarative, no-code data dashboards with Lumen YAML specifications
• •lumen-ai - AI-powered natural language data exploration with Lumen AI

When to use: Interactive dashboards, web applications, large-scale data visualization, geographic mapping, real-time data streaming, exploratory data analysis, publication-quality visualizations

GAIA Data Downloader Plugin

Generate and develop hydroclimatological data download scripts for the GAIA project, covering 10+ environmental data sources.

Agent:

• •Data Downloader - Generates download scripts from natural language prompts using a three-phase interaction (propose config → confirm → generate) for CONUS404, HRRR, WRF, PRISM, Stage IV, USGS, ORNL, DEM, Synoptic, and IRIS data

Skill:

• •download-script-dev - Templates, configuration validation, and debugging guidance for developing data download scripts

When to use: Downloading hydroclimatological data for GAIA geoscience research, creating reproducible data pipelines, setting up new data download workflows

Browse the plugins directory and community-plugins directory to explore all available plugins.

Repository Structure

rse-plugins/
├── .claude-plugin/
│   └── marketplace.json                                # Claude plugin marketplace configuration
├── plugins/                                            # Main plugin collection
│   ├── scientific-python-development/                  # Scientific Python development plugin
│   │   ├── .claude-plugin/
│   │   │   └── plugin.json
│   │   ├── agents/
│   │   │   ├── scientific-python-expert.md
│   │   │   └── scientific-docs-architect.md
│   │   └── skills/
│   │       ├── pixi-package-manager/
│   │       ├── python-packaging/
│   │       ├── python-testing/
│   │       ├── code-quality-tools/
│   │       └── scientific-documentation/
│   ├── scientific-domain-applications/                 # Domain-specific scientific computing plugin
│   │   ├── .claude-plugin/
│   │   │   └── plugin.json
│   │   ├── agents/
│   │   │   └── astronomy-astrophysics-expert.md
│   │   └── skills/
│   │       ├── xarray-for-multidimensional-data/
│   │       └── astropy-fundamentals/
│   ├── ai-research-workflows/                          # AI-enabled research workflow plugin
│   │   ├── .claude-plugin/
│   │   │   └── plugin.json
│   │   ├── agents/
│   │   │   └── research-workflow-orchestrator.md
│   │   ├── commands/
│   │   │   ├── research.md
│   │   │   ├── plan.md
│   │   │   ├── iterate-plan.md
│   │   │   ├── experiment.md
│   │   │   ├── implement.md
│   │   │   └── validate.md
│   │   └── skills/
│   │       └── research-workflow-management/
│   │           ├── SKILL.md
│   │           └── assets/
│   │               ├── research-template.md
│   │               ├── plan-template.md
│   │               ├── experiment-template.md
│   │               └── implement-template.md
│   └── project-management/                              # Project lifecycle management plugin
│       ├── .claude-plugin/
│       │   └── plugin.json
│       ├── agents/
│       │   ├── project-onboarding-specialist.md
│       │   └── documentation-validator.md
│       ├── commands/
│       │   ├── setup-project.md
│       │   ├── project-handoff.md
│       │   └── validate-project-handoff.md
│       └── skills/
│           ├── community-health-files/
│           └── documentation-validation/
├── community-plugins/                                  # Community-contributed plugins
│   ├── gaia-data-downloader/                           # GAIA data download plugin
│   │   ├── .claude-plugin/
│   │   │   └── plugin.json
│   │   ├── agents/
│   │   │   └── data-downloader.md
│   │   ├── skills/
│   │   │   └── download-script-dev/
│   │   │       ├── SKILL.md
│   │   │       └── references/
│   │   ├── AGENTS.md
│   │   ├── CLAUDE.md
│   │   └── README.md
│   └── holoviz-visualization/                          # HoloViz ecosystem plugin
│       ├── .claude-plugin/
│       │   └── plugin.json
│       ├── agents/
│       │   ├── panel-specialist.md
│       │   ├── visualization-designer.md
│       │   ├── data-engineer.md
│       │   └── geo-spatial-expert.md
│       ├── skills/
│       │   ├── panel-dashboards/
│       │   ├── plotting-fundamentals/
│       │   ├── data-visualization/
│       │   ├── advanced-rendering/
│       │   ├── geospatial-visualization/
│       │   ├── colormaps-styling/
│       │   ├── parameterization/
│       │   ├── lumen-dashboards/
│       │   └── lumen-ai/
│       └── references/                                 # HoloViz ecosystem documentation
│           ├── holoviz-ecosystem.md
│           ├── library-matrix.md
│           ├── best-practices/
│           ├── patterns/
│           ├── troubleshooting/
│           ├── lumen-dashboards/
│           ├── lumen-ai/
│           └── colormaps/
├── CONTRIBUTING.md                                     # Contribution guidelines
├── LICENSE                                             # BSD 3-Clause License
└── README.md                                           # This file

Architecture

Plugin System

This repository uses the Claude Code plugin marketplace architecture:

• •Plugins - Top-level containers organized by domain (e.g., python-development, scientific-computing)
• •Agents - Specialized AI personas with comprehensive expertise in specific areas
• •Skills - Reusable knowledge modules that provide detailed guidance on specific topics

Design Philosophy

The agents and skills follow the Scientific Python Development Guide principles:

1. Collaborate - Adopt conventions and tooling used by the broader scientific Python community
2. Refactor Fearlessly - Leverage tests and tools to enable confident iteration
3. Prefer Wide Over Deep - Build reusable, extensible solutions for unforeseen applications

Agent vs Skill

• •Agents - Comprehensive personas that handle complete workflows, make decisions, and provide end-to-end guidance
• •Skills - Focused knowledge modules on specific topics (e.g., testing patterns, packaging workflows) that agents can reference

Contributing

We welcome contributions of new agents, skills, and improvements! Please see CONTRIBUTING.md for guidelines on:

• •Creating new agents and skills
• •Plugin organization and structure
• •Naming conventions and best practices
• •Testing and validation
• •Submitting pull requests

Documentation

For detailed information about the plugins and their contents:

• •Contributing Guidelines - How to contribute to this repository
• •HoloViz Ecosystem Overview - Introduction to the HoloViz ecosystem
• •HoloViz Library Matrix - Comparison of HoloViz libraries and when to use each

Related Resources

Scientific Python Community

• •Scientific Python Development Guide - Community best practices
• •Scientific Python Lectures - Educational materials
• •NumPy, SciPy, Pandas - Core libraries

HoloViz Ecosystem

• •HoloViz.org - Main HoloViz ecosystem portal
• •Panel - Build interactive dashboards and web applications
• •hvPlot - High-level plotting API for pandas and xarray
• •HoloViews - Declarative data visualization
• •Datashader - Render large datasets accurately
• •GeoViews - Geographic data visualization
• •Lumen - No-code dashboards with AI capabilities
• •Param - Declarative parameter management
• •Colorcet - Perceptually uniform colormaps

Domain-Specific Libraries

• •Astropy - Astronomy and astrophysics in Python
• •Xarray - Labeled multidimensional arrays for climate and Earth science

Research Software Engineering

• •UW Scientific Software Engineering Center
• •Best Practices for Scientific Computing
• •The Turing Way - Guide to reproducible research

Claude Code

• •Claude Code Documentation
• •Claude Plugin Marketplace

License

This project is licensed under the BSD 3-Clause License - see the LICENSE file for details.

Acknowledgments

Developed and maintained by the University of Washington Scientific Software Engineering Center (UW-SSEC).

Questions or Issues?

Please open an issue on GitHub.