nvoxland

Persistent, versioned, queryable data layer. Use when the user wants to analyze CSV/Parquet/JSON files, explore Kaggle datasets, clean or transform data, join multiple data sources, build reusable datasets, share data with a team, version data changes, create data pipelines, fetch data from APIs or cloud storage, or do any multi-step data work that should persist.

Use oxplow like a real user to get work done while QA/design-reviewing it — drive the in-oxplow agent as the navigator, file bugs as oxplow tasks, document findings in ideas/qa. Use when the user says "dogfood oxplow", "use oxplow as a user", "navigator session", or asks for a QA/design-review session of oxplow itself.