weather‑skills.org 41 skills MIT license github

An AI-ready toolkit for weather and climate data

Use chat to fetch, transform, and visualize weather data — with provenance you can audit

Weather skills are composable tools that allow AI agents to support operational forecasting, scientific exploration of climate data, and the use of forecasts and weather data for specific applications. Each weather skill is expert-reviewed, and its outputs are generated deterministically by that reviewed code — the agent only decides which tools to call, it never touches the underlying data. Every result carries provenance, so you can inspect the exact chain of tool calls that produced it. Initiated by Rhiza Research, with a community effort to steward the catalog.

Request access → request early access to the hosted weather skills platform
weather skills
  1. Fetch forecasts and observations
  2. Transform clip, aggregate, convert
  3. Visualize maps and time series

Provenance is recorded at every step.

01

See it run

From a natural-language request, the agent picks skills and runs them in order. This is a real session, replayed.

agent session
02

Getting started

Skills are simple, expert-reviewed, Python scripts with descriptions that assist AI agents in calling those scripts. Use them through our hosted chat interface, via MCP, in your favorite agent, or directly in your terminal.

We are rolling out a hosted version of weather skills. Request access if you are interested in being a beta tester.

Screen recording of Weather Skills Chat. The Weather Agent offers prompts to show observations, download a forecast, or compare a forecast with observations.

Request access →

A hosted MCP will be coming soon.

For use by a local agent, install the SKILL.md files into your project with skillkit:

# List what skillkit discovers in the repo
npx skillkit install weather-skills/weather-skills-catalog --list

# Install all skills to the current project
npx skillkit install weather-skills/weather-skills-catalog --all --yes

# Install just a subset
npx skillkit install weather-skills/weather-skills-catalog --skill=ecmwf-fetch

For command-line use, with no install:

# List available skills
uvx --from git+https://github.com/weather-skills/weather-skills-catalog weather-skills

# Run one
uvx --from git+https://github.com/weather-skills/weather-skills-catalog weather-skills <skill> [args]
03

Adding your own skills

Do you have a new forecasting model, downscaling method, or weather data visualization? Are you an expert in a field that relies on weather data like disaster early warning, energy, or agriculture? We are excited to work with you to build skills to access your data or enable your use case!

To make sure skills work well together, skills follow standard data format and argument conventions, but most of the work is handled by the weather-skills-core package so that you can focus on your application.

Check out existing skills in the weather skills catalog to get started, or reach out to us at help@weather-skills.org so that we can help.

Submit a skill →

04

Skill catalog

A view of current capabilities. Click a skill for a short description. Full write-ups are in the weather skills catalog.

Datasources

  • AIFS

    ECMWF Artificial Intelligence Forecasting System, via the dynamical.org catalog.

  • CHIRPS

    UC Santa Barbara Climate Hazards Center precipitation.

  • CMIP6

    CMIP6 climate-model output from the Pangeo catalog on Google Cloud.

  • ECMWF S2S

    ECMWF subseasonal-to-seasonal ensemble, from the ECMWF Data Stores.

  • ERA5

    ECMWF ERA5 reanalysis, from the ARCO store on Google Cloud.

  • GEFS

    NOAA Global Ensemble Forecast System, via the dynamical.org catalog.

  • GFS

    NOAA Global Forecast System, via the dynamical.org catalog.

  • GHCN

    NOAA Global Historical Climatology Network daily station observations.

  • ICON-EU

    DWD ICON-EU regional forecast, via the dynamical.org catalog.

  • IFS-ENS

    ECMWF Integrated Forecasting System ensemble, via the dynamical.org catalog.

  • IMERG

    NASA IMERG satellite precipitation, from NASA Earthdata.

  • MRMS

    NOAA Multi-Radar Multi-Sensor precipitation analysis, via the dynamical.org catalog.

  • OISST

    NOAA Optimum Interpolation sea-surface temperature, from NOAA Physical Sciences Laboratory.

  • OpenAQ

    Air-quality station observations from the OpenAQ network.

  • SMAP

    NASA SMAP soil moisture, from NASA Earthdata.

  • SubC MME

    Climate Hazards Center SubC multi-model ensemble.

  • TAHMO

    TAHMO weather stations across Africa.

Transformations

  • aggregate

    Roll a time series up into daily, weekly, dekadal, or monthly windows.

  • calendar

    Convert a dataset's time axis onto another calendar.

  • clip

    Cut a dataset down to a bounding box or polygon.

  • coarsen

    Regrid a dataset onto a coarser or realigned grid.

  • concat

    Join datasets along one dimension.

  • deaccumulate

    Turn a cumulative forecast into a per-step rate.

  • difference

    Subtract one dataset from another, cell by cell.

  • downscale

    Map a dataset onto a finer grid.

  • IOD

    Compute the Indian Ocean Dipole index from a temperature anomaly.

  • rename

    Rename one variable in a dataset.

  • select

    Keep chosen entries along one dimension.

  • step to time

    Turn forecast lead times into calendar valid times.

  • summarize

    Collapse a dimension with a statistic such as the mean or the spread.

  • totals

    Convert a rate into a total over its aggregation period.

  • units

    Convert variables into different units.

Visualizations

  • heatmap

    Draw a map or a single time series from one dataset.

  • heatmap-compare

    Compare datasets as heatmaps, side by side or across times.

  • ITF

    Show the latest NOAA CPC map of the African Intertropical Front.

  • mediogram

    Compare a forecast ensemble with its historical climate at one place.

  • MJO

    Show the latest NOAA CPC Madden–Julian Oscillation phase diagram.

  • timeseries

    Overlay several datasets as lines on one time axis.

Utilities

  • feedback

    Build a link that files a GitHub issue about the skills.

  • inspect data

    List a dataset's dimensions, coordinates, and variables.

  • provenance

    Show how an output was produced and how to regenerate it.

  • resolve-region

    Turn a place name into a bounding box or a boundary.

  • resolve-time

    Turn a relative date into absolute start and end times.

05

Governance

The library will be overseen by a working group of weather and climate data practitioners. A submission and review process will define how new skills are proposed, tested against the standard dataset contract, and admitted to the catalog.

Until that process is in place, contributions go through the repository directly: open an issue or a pull request.