soil

soil-skills

Lightweight Claude skills for live SSURGO soil data. No R, no soilDB, no installs — just markdown that teaches your agent two keyless public APIs.

Skills

soil-survey

Give it a US address (or coordinates) and get a readable soil report: what soil is there, its profile and properties, what USDA says works on it (septic, building, lawns), farmland ratings, and water/flooding hazards — all queried live from the USDA soil survey (SSURGO) via Soil Data Access.

> what's the soil like at 1024 Olive Dr, Davis CA? could it support a vineyard?

It also handles:

• Area / parcel — paste a WKT or GeoJSON polygon (or give a radius) for area-weighted soils across a parcel, not just one point.
• Compare — soils across two or more locations, side by side.
• Series lookup — "tell me about the Yolo series" (by name, no address).

soil-carbon

Estimate soil carbon stock at any US location — organic (SOC) and inorganic (SIC, carbonates) — in tonnes C/ha at 0–30 cm and 0–100 cm, computed transparently from SSURGO horizon data (organic matter, carbonate, bulk density, thickness, coarse-fragment correction), each with an uncertainty band from SSURGO's low/high estimates and the method shown. Works for a single point or an area/parcel (pasted WKT/GeoJSON polygon or a radius) — area-weighted mean and total tonnes.

> how much carbon is in the soil at 1024 Olive Dr, Davis CA?

Install

As a Claude Code plugin (gets both skills):

/plugin marketplace add clevinson/soil-skills
/plugin install soil@soil-skills

Then ask in plain language, or invoke explicitly: /soil:soil-survey <address> · /soil:soil-carbon <address>.

Via the skills CLI (works across Claude Code, ChatGPT, Codex, and other agents that support the Agent Skills standard):

npx skills add clevinson/soil-skills@soil-survey
npx skills add clevinson/soil-skills@soil-carbon

Network access: these are external APIs. Sandboxed runtimes (ChatGPT, claude.ai) block outbound internet by default — allowlist geocoding.geo.census.gov and sdmdataaccess.sc.egov.usda.gov, then start a fresh conversation. Claude Code's shell is generally open.

How it works

The skills are pure markdown. At runtime the agent:

1. Geocodes the address (Census Bureau geocoder, keyless)
2. Finds the SSURGO map unit for the point (USDA Soil Data Access, T-SQL over HTTPS)
3. Pulls map unit, component, horizon, and interpretation data with pre-tested SQL templates
4. Writes the report — every number traceable to a live query response (soil-carbon additionally shows the SOC formula applied to those values)

License & data

Code in this repo is licensed under the MIT License.

The soil data itself is not covered by that license and isn't redistributed here — it's queried live from US federal government sources, which are in the public domain (17 U.S.C. § 105):

• Soil data: USDA Natural Resources Conservation Service, Soil Survey Geographic Database (SSURGO), via Soil Data Access.
• Geocoding: US Census Bureau Geocoding Services.

USDA requests that SSURGO data be cited when used; please keep the attribution above if you build on this.