An agent organization structured as a company — 15+ departments, 125+ skills, each independently installable, citing the standards and regulators that settle the question. Runs in Claude Code and ChatGPT.
Sources — data-analytics:data-engineering
Check these before answering on anything they cover, and cite what you used. The use note on each one is binding: most of what a professional cites is free to read and not free to reproduce.
Apache Iceberg table specification
Apache Software Foundation · global · free to use with attribution — credit the publisher
https://iceberg.apache.org/spec/
Authoritative for: The on-disk contract — metadata layout, snapshots, schema and partition evolution, and what isolation a reader may assume. The document that decides whether two engines agree on what a table is.
Apache Parquet file format specification
Apache Software Foundation · global · free to use with attribution — credit the publisher
https://parquet.apache.org/docs/file-format/
Authoritative for: Physical layout, encodings, nesting and the statistics a reader may rely on for predicate pushdown. The normative answer to why a file is slow or unreadable.
Data Catalog Vocabulary
World Wide Web Consortium · global · read and cite only — copyrighted, do not reproduce
https://www.w3.org/TR/vocab-dcat-3/
Authoritative for: The interoperable vocabulary for describing datasets and catalogs — what a catalog entry must carry to be exchangeable rather than specific to one tool.
OpenLineage object model
LF AI and Data Foundation · global · free to use with attribution — credit the publisher
https://openlineage.io/docs/spec/object-model/
Authoritative for: The vendor-neutral event schema for lineage — what an event must contain to be portable, rather than whatever one catalog happens to record.
Shapes Constraint Language
World Wide Web Consortium · global · read and cite only — copyrighted, do not reproduce
Authoritative for: How a data constraint is expressed and what conformance means — the normative answer to whether a record is valid against a model, rather than each tool inventing its own semantics.
Sources are maintained in sources/ upstream, not here. If one is wrong, out of date, or missing, fix it there — this file is regenerated and an edit to it is lost.