A shared, living collection of skill specifications for the Denolle Lab at the University of Washington. We do seismology, geophysics, and AI-driven hazard science — and we use AI tools daily. This repo captures how we work, so that knowledge is discussable, transferable, and improvable.
A skill file is a structured document that makes domain knowledge explicit. Two types live here:
| Type | Format | Purpose |
|---|---|---|
| AI skill | TEMPLATE-ai-skill.md |
Configure a Copilot/LLM agent for a specific lab task — knowledge anchor, standing instructions, prompt intent |
| Research method | TEMPLATE-method.md |
Document a research workflow, protocol, or technique so it can be learned, reused, and critiqued |
Both types are designed to be discussable. Use GitHub Discussions to propose changes, ask questions, or share what worked.
Caution Note: None of these skill files has been rigorously evaluated.
ai-tools/ AI/Copilot skill files for lab-specific tasks
research-methods/ Seismology and geophysics method docs
data-analysis/ Python workflows and data processing guides
writing/ Paper, grant, and scientific communication guides
AI skills — load the file as a Copilot agent instruction or paste the standing instructions block into your preferred LLM. Upload the listed knowledge anchor document types before prompting.
Research methods — read as a protocol. Follow the workflow steps, adapt to your dataset, and note any deviations in your lab notebook.
See CONTRIBUTING.md for the full guide. Short version:
- Open a new-skill issue or improve-skill issue to start the conversation
- Fork → branch (
skill/short-name) → PR - Follow the appropriate template — one of
TEMPLATE-ai-skill.mdorTEMPLATE-method.md - Include at least one honest failure mode before submitting
- academic-practice-agents — broader role-based agent framework for faculty academic life (the general framework this repo builds on)
- Denolle Lab website — Skills page linking to this repo
MIT — use, adapt, and share freely with attribution.