An advanced, production-grade AI agent skill pack, custom agent directory, and Model Context Protocol (MCP) specification framework designed for astronomical research, literature synthesis, and statistical inference workflows.
This toolkit is currently optimised for pulsar astronomy, pulsar timing array (PTA) noise modeling, Bayesian parameter estimation, and reproducible scientific computing. It is designed to act as a Research Copilot for Pulsar Astronomers, installable into AI assistants such as Claude Code, Cursor, Cline, Roo Code, and other MCP-enabled LLM environments.
graph TD
User([Astronomer / User]) -->|Interacts with| Agent[Custom Astronomy Agent]
Agent -->|Orchestrates| Workflows[Research Workflows]
Agent -->|Invokes| Skills[Specialised AI Skills]
Agent -->|Accesses Data & Tools via| MCP[Model Context Protocol Servers]
subgraph Skills Directory
Skills --> LitSkills[Literature Analysis]
Skills --> InfSkills[Inference: Emcee, Bilby, Enterprise]
Skills --> VisSkills[Publication-Grade Visualisation]
Skills --> SESkills[Scientific Software Engineering]
end
subgraph MCP Integration
MCP --> ADS[NASA ADS / arXiv]
MCP --> SIMBAD[SIMBAD / Vizier]
MCP --> Archives[MAST / NRAO / ESA / SKA]
MCP --> Local[FITS / AstroPy / CASA / Jupyter]
end
- Claude Code: Full support for skills, workflows, and agents. Configured via the
.claude-plugin/metadata files. - MCP Clients (Cursor, Cline, Roo Code) (currently in progress): Integrates the MCP specifications defined under
mcps/to enable secure data retrieval and computation. - Open WebUI: Importable JSON prompts and agents using templates and system prompts.
To load these skills and configurations into your Claude Code environment, clone this repository and configure your assistant path:
git clone https://github.com/astronomy-collaboration/astronomy-ai-toolkit.git
# Link the plugin configuration
cp -r astronomy-ai-toolkit/.claude-plugin ~/.claude-pluginskills/: The core system prompts and skill manuals for 31 specific capabilities.agents/: Orchestrator agent definitions (literature, proposals, observation planning, data reduction, transients, supervisor).mcps/: Architectural specifications and tool schemas for 13 astronomy-related Model Context Protocol servers.workflows/: Reusable multi-step execution plans for complex research tasks.templates/: Markdown templates for reports, checklists, and summary logs.examples/: Step-by-step guides using real pulsar astronomy scenarios and mathematics..claude-plugin/: Integration metadata (marketplace.json,plugin.json) for agent clients.
- [] Fix MCP Folder to ensure consistent format
- [] Fix examples folder to ensure examples meet expected formats used
- [] Make sure workflow folder meet expected formats used
- [] Add more Python-related skills