The lab keeps its AI tooling in one place: wilson-lab-ai-skills. If you use Claude to help with lab work, start there.
What it is#
A shared repo of things that make an AI agent genuinely useful for our work:
- skills/ — reusable “skills” that teach Claude a lab workflow once, so nobody has to re-explain it. Examples: reading long PDFs, putting files on the lab server, keeping this wiki tidy.
- tutorials/ — slides from lab-meeting intros to agentic AI.
- mcp/ — notes on connectors (MCP servers) worth plugging in.
- CLAUDE.md — standing instructions agents follow when working in the repo.
How to use a skill#
You don’t “run” a skill. Open the repo in Claude Code or Cowork, and skills load themselves when your request matches — you just ask for the thing. For example, “read this thesis and summarise it” pulls in the PDF skill; “put these slides on the lab server” pulls in the server skill. If nothing matches, Claude just works normally.
Start here#
New to this? Open tutorials/agentic_ai_labmeeting.pptx — the lab’s gentle introduction to
agentic AI, git, and the Claude toolkit. Then read the
README.
Add your own#
Found a workflow you keep re-explaining? That’s a skill waiting to happen. Ask Claude to use the making-skills skill and it will help you write one in the house format. Contribute it back with a pull request so the rest of us benefit.
This page was drafted with AI assistance — fittingly, since it’s about using AI.