Datasets
The data skills are evaluated on, in the open. Every dataset carries a schema, a license and its provenance, including whether consent was recorded. A dataset with consent pending is never listed. Creator text is a dataset too: kind text, one row per post, and when the text is licensed a signed consent record beside the manifest that says what was licensed (text only), for which uses, on what revenue share and revocation terms.
| Dataset | Kind | Rows | Size | Origin / consent | License | Used by |
|---|---|---|---|---|---|---|
| demo-creator-maya Sixty synthetic posts by an invented outdoors creator, in one consistent voice, standing in for a real creator's licensed export. | text | 60 | 20.0 KB | synthetic consent not applicable | CC0-1.0 | |
| demo-meta-campaigns Weekly spend, clicks, purchases and attributed revenue per campaign for the Casper demo Meta ad account, joinable to the demo store by product title. | tabular | 78 | 21.0 KB | synthetic consent not applicable | CC0-1.0 | |
| demo-store-catalog Products with price, unit cost and gross margin, plus variant inventory, for the Casper demo store. | tabular | 14 | 1.1 KB | synthetic consent not applicable | CC0-1.0 | |
| demo-store-orders 90 days of orders from the Casper demo store: the deterministic fake store every skill eval on the hub runs against. | tabular | 2,769 | 819.8 KB | synthetic consent not applicable | CC0-1.0 |
Add one with pnpm casper dataset init datasets/my-data. Validation checks every row against the declared schema.