# Add a discipline or improve a collection

1. Choose an existing folder or a new English subject slug. Use the [OpenAlex mapping](taxonomy/openalex_topics.csv) to plan coverage when helpful. Add a definition to `collections.json` for a new subject.
2. Read [Best-Practices.txt](Best-Practices.txt) and give a worker [SEARCH_AGENT_PROMPT.txt](SEARCH_AGENT_PROMPT.txt) plus `DOMAIN: <your discipline>`. Coordinate teams using [ORCHESTRATOR_GUIDE.txt](ORCHESTRATOR_GUIDE.txt).
3. Search normal public academic sources and download PDFs into private storage outside the publication checkout. Verify complete-file access and inspect content. Keep real query logs and rejected candidates privately.
4. Add sanitized accepted records to `data/verified_links.jsonl` using [SCHEMA.md](SCHEMA.md). Keep unresolved leads in `data/legacy_candidates.jsonl` with their explicit unverified status. Never include PDFs or extracted text.
5. Run `python scripts/catalog.py render .` and `python scripts/catalog.py validate .`. Review the resulting Markdown, counts and README examples. Update root release counts when records change.
6. Propose the change through Hugging Face Discussions or the dataset's contribution interface. Describe the scope, check date and review evidence. Publication remains a maintainer action.

Prefer a stable source page plus a usable public PDF-entry link. Keep exact editions and original work titles. Do not replace a long textbook with a short summary or present a portal as a full-text result. The repository license applies to the collection, not to linked documents.