curate-next

Batch curation of uncatalogued stub resources is done by selecting eligible candidates from your local repository and dispatching per-resource curation in parallel. You can curate the next N stubs by asking your agent to process a specific batch size.

To skip previously blocked stubs during data curation, the workflow excludes anything already recorded in reports/curation_problems.tsv. This prevents re-curating already logged problems and ensures iterative content improvement workflows avoid known issues.

Parallel agent curation prevents conflicting edits by running the kg-registry-curation workflow for each selected resource with strict isolation. This isolated parallel execution ensures multiple resources are curated simultaneously without overwriting each other's changes.

You validate curated KG-Registry files after batch processing by running local make-based checks. The workflow ensures each edited resource passes 'uv run make validate-file FILE=...', with re-validation after prettification if needed to confirm structural integrity.

KG-Registry provenance resolution requires specific identifiers to prevent common provenance mistakes. Workers must use KG-Registry identifiers for original_source and secondary_source, and mint new unique KG-Registry IDs when upstream resources are not yet present in the registry.

Strict input validation for N is required before dispatching per-resource curation to ensure the batch size is valid. This validation step is mandatory before the workflow selects uncatalogued stub candidates and begins parallel processing.