AgentSkillsCN

essdive-datasets

搜索ESS-DIVE数据集，获取元数据与权限信息，并通过MCP工具解析FLMD。

SKILL.md
--- frontmatter
name: essdive-datasets
description: Search ESS-DIVE datasets, fetch metadata/permissions, and parse FLMD via MCP tools.

Setup (once)

Run the MCP server locally:

bash
uv run python src/essdive_mcp/main.py --token YOUR_ESS_DIVE_TOKEN_HERE

Or provide the token via a file:

bash
uv run python src/essdive_mcp/main.py --token-file /path/to/token.txt

Note: the environment variable name is ESSDIVE_API_TOKEN (no underscore between ESS and DIVE).

Register with Claude Code (stdio transport):

bash
claude mcp add --transport stdio essdive-mcp --env ESSDIVE_API_TOKEN=YOUR_ESS_DIVE_TOKEN_HERE -- uv run python ./src/essdive_mcp/main.py

Token file alternative:

bash
claude mcp add --transport stdio essdive-mcp -- uv run python ./src/essdive_mcp/main.py --token-file /path/to/token.txt

Tools

  • •search-datasets
  • •get-dataset
  • •get-dataset-permissions
  • •parse-flmd-file
  • •coords-to-map-links

Usage examples

Search datasets by keyword:

code
search-datasets with query="soil carbon" and page_size=10

Filter by creator and provider:

code
search-datasets with creator="Smith" and provider_name="NGEE Arctic" and page_size=5

Get detailed metadata for a dataset:

code
get-dataset with id="ess-dive-9ea5fe57db73c90-20241024T093714082510"

Check sharing permissions (requires token):

code
get-dataset-permissions with id="ess-dive-9ea5fe57db73c90-20241024T093714082510"

Parse a File Level Metadata CSV payload:

code
parse-flmd-file with content="filename,file_description\nfile1.csv,Soil moisture data\n"

Create map links for a point (geojson.io + Google Maps + Google Earth KML):

code
coords-to-map-links with points=[[38.9219, -106.9490]] and zoom=12

Create map links for a bounding box (geojson.io + OpenStreetMap + Google Earth KML):

code
coords-to-map-links with bbox=[38.9187, -106.9532, 38.9263, -106.9451]

Notes

  • •Use format="summary" for compact results, or format="detailed" for full metadata.
  • •page_size max is 100; row_start is 1-based.

Fallback (no MCP server)

Search via the ESS-DIVE API:

bash
curl -sG "https://api.ess-dive.lbl.gov/packages" \
  -H "Accept: application/json" \
  -H "User-Agent: Mozilla/5.0" \
  -H "Range: bytes=0-1000" \
  --data-urlencode "text=soil carbon" \
  --data-urlencode "page_size=10" \
  --data-urlencode "row_start=1" \
  --data-urlencode "isPublic=true"

Fetch a dataset record by package ID:

bash
curl -sG "https://api.ess-dive.lbl.gov/packages/REPLACE_WITH_PACKAGE_ID" \
  -H "Accept: application/json" \
  -H "User-Agent: Mozilla/5.0" \
  -H "Range: bytes=0-1000" \
  --data-urlencode "isPublic=true"