.ch$

1. A new GitHub issue is opened in this repo with your form contents in a structured YAML block and the label ext-review. Your GitHub login becomes the annotator; the issue's created_at is the submission timestamp.
2. A curator script (tools/process_extension_labels.py) fetches all ext-review issues and writes their parsed contents into data/derived/extension_labels.csv with curator_status="new". Until this script runs, your submission lives only as the GitHub issue.
3. Your submission shows up on the curator triage page (/review/curator/) under "new". Maintainers review the issue (comment, ask for clarification, accept or reject) and edit extension_labels.csv to set curator_status to accepted.
4. For pl/<id> labels, build_pl_taxonomy.py promotes accepted labels into ext_claim.csv with source="manual_review:<your-login>" and strength="proposed". The next site rebuild then shows your claim on this extension's page and on the language's page.
5. Friendly-name + reference-URL labels (e.g. binary:image for .png) are not promoted into ext_claim.csv (they aren't PL claims) but are surfaced on this page's "Prior labels" table so anyone landing here sees them next to the auto-detected category.

In short: submitting opens an issue (provenance), and a small ingestion pipeline turns issues into rows the site renders. The loop is currently manual (the curator script must be run by a maintainer); GitHub Actions could automate it on each new issue.