earth-space-ai.org is a collection of skill packages for Earth, planetary, and space science models.Each repo is a self-contained, progressive-disclosure knowledge package: a SKILL.md routing hub plus reference/ deep-dive docs covering install, compile, run, modify, debug, and contribute.Designed to be loaded as skills by AI coding agents (Claude Code, Codex, Cursor) and to serve as durable, human-readable references for researchers and developers.
Mechanistic Earth-system models (CESM, E3SM, WRF, MOM6, Noah-MP, CTSM, JULES, ...) carry decades of accumulated scientific judgment in their source trees, build systems, and debug folklore. Most of that knowledge lives in PDFs, mailing lists, and the heads of senior researchers.
A skill repo is a structured, progressive-disclosure package of that knowledge: a routing hub (SKILL.md) for top-level intent, and a reference/ tree of deep-dive docs by topic (architecture, physics, workflow, debugging, ...). It is designed to be loaded into an agent's context on demand and to stay readable as a human reference.
Each repo carries its own license. Each is maintained alongside the model it covers. We currently list 32 skills and resources across Earth-system, atmosphere, land, ocean, sea-ice, solid Earth, and heliophysics domains.
A profile of benchmark design, physics-aware code assistance, scientific skill extraction, and daily AI-assisted research workflows. Responding editor: Zesen Huang.
How evaluation, agent workflows, and scientific modeling meet in one research practice.
Open articleStatus legend: deep-dive. scaffold, routing and source-grounded surface verified, operational depth being filled in.
Open an issue or PR on the relevant repo. For new model coverage that fits the scope (Earth, ocean, atmosphere, land, ice, space weather, planetary), open an issue on the org before starting a new repo.
Contact: zesenhuang@g.ucla.edu; ktwu@utexas.edu