name

source-data-package

description

Set up and manage Frictionless Data Package repos created from the recallnet source-data-package-template. Use this skill whenever the user is working in a repo created from the source-data-package-template, mentions adding source data, creating a data package, setting up a new data ingestion repo, registering data resources, or working with datapackage.json. Also use it when the user asks about LFS setup for data repos, date-folder conventions, or Frictionless Data Package structure in the context of recallnet repos.

Source Data Package

This skill guides you through setting up and working with repos created from the recallnet/source-data-package-template. These repos follow the Frictionless Data Package spec and use Git LFS for data files.

Repo structure

datapackage.json          # package descriptor (metadata + resource list)
.gitattributes            # LFS tracking rules for data/**
data/                     # date-stamped folders for raw data (LFS-tracked)
  YYYY-MM-DD/
    source-file.csv
scripts/                  # preparation, transformation, or validation scripts
README.md

First-time machine setup

Git LFS must be installed and configured before cloning or pushing data. Without it, data files won't be stored correctly and clones will get pointer files instead of real data.

Install git-lfs — on nix-darwin managed machines, add git-lfs to home.packages in home.nix and rebuild. On other systems: brew install git-lfs or apt install git-lfs.

Configure LFS globally — either run git lfs install or (on nix-darwin) add the filter config to programs.git.settings:

filter.lfs.clean = "git-lfs clean -- %f";
filter.lfs.smudge = "git-lfs smudge -- %f";
filter.lfs.process = "git-lfs filter-process";
filter.lfs.required = true;

Verify — git lfs version should print a version string and git config filter.lfs.required should return true.

Setting up a new data package repo

After creating a repo from the template:

1. Update the descriptor

Edit datapackage.json and replace the placeholder values:

name — lowercase, url-safe identifier (alphanumeric, ., _, - only). Use the repo name.
title — human-readable title for the dataset.
description — what this data package contains and why it exists.
id — a unique identifier (e.g., a UUID or a doi). Can be left blank initially.
sources — where the data comes from. Fill in title (source name) and path (URL to the source).
licenses — defaults to CC-BY-4.0. Change if the data has different licensing.
version — starts at 0.1.0. Bump as the package evolves.

Example:

{
  "name": "coingecko-market-data",
  "id": "",
  "title": "CoinGecko Market Data",
  "description": "Daily market snapshots from the CoinGecko API",
  "version": "0.1.0",
  "licenses": [
    {
      "name": "CC-BY-4.0",
      "path": "https://creativecommons.org/licenses/by/4.0/",
      "title": "Creative Commons Attribution 4.0 International"
    }
  ],
  "sources": [
    {
      "title": "CoinGecko API",
      "path": "https://api.coingecko.com/api/v3"
    }
  ],
  "resources": []
}

2. Add data files

Place data files in date-stamped folders under data/:

data/2026-03-19/market-snapshot.csv
data/2026-03-19/token-metadata.json

The date represents when the data was captured or ingested, not when it was published at the source. Use ISO 8601 format (YYYY-MM-DD).

All common data formats placed under data/ are automatically tracked by Git LFS via .gitattributes. The tracked extensions are: csv, tsv, json, jsonl, parquet, arrow, avro, xlsx, xls, sqlite, db, gz, zip, tar, bz2.

If you need to add a format not in this list, append a line to .gitattributes:

data/**/*.newext filter=lfs diff=lfs merge=lfs -text

3. Register resources

Every data file should be registered as a resource in datapackage.json. Add an entry to the resources array for each file:

{
  "name": "market-snapshot",
  "path": "data/2026-03-19/market-snapshot.csv",
  "format": "csv",
  "mediatype": "text/csv",
  "description": "End-of-day market snapshot with price, volume, and market cap"
}

Resource fields:

name — lowercase identifier for this resource (same naming rules as the package name)
path — relative path to the file from the repo root
format — file extension (csv, json, parquet, etc.)
mediatype — MIME type (text/csv, application/json, application/vnd.apache.parquet, etc.)
description — what this specific resource contains

When multiple date folders contain the same logical resource (e.g., daily snapshots), register each file as a separate resource or use the most recent one and update the path on each ingestion — whichever convention the team agrees on.

4. Add scripts (optional)

Place any data preparation, transformation, or validation scripts in scripts/. These might include:

Fetch scripts that pull data from an API
Transformation scripts that clean or reshape raw data
Validation scripts that check data quality

Common operations

Adding a new day's data

mkdir -p data/$(date +%Y-%m-%d)
# copy or fetch data into the new folder
# update resources in datapackage.json
git add data/$(date +%Y-%m-%d)/ datapackage.json
git commit -m "data: add $(date +%Y-%m-%d) snapshot"

Verifying LFS is tracking correctly

git lfs ls-files          # list LFS-tracked files in the repo
git lfs status            # show pending LFS uploads

If a data file shows up in git diff as full content instead of a pointer, LFS isn't tracking it — check .gitattributes and ensure git lfs install has been run.