kg-registry-validation

kg-registry-validation

Validate a KG-Registry resource page after editing, with extra attention to provenance and curation-specific quality issues.

Prerequisite

• Run this skill only from a local clone of the KG-Registry repository.
• The workflow depends on validating local files in resource/ and running repository-local commands such as uv run make validate-file.
• If the repository is not cloned locally, stop and clone it before attempting validation.

When to use

• After curating or expanding a resource page.
• When a resource file fails validate-file.
• When reviewing whether original_source and secondary_source were filled correctly.
• When checking that a stub was fully converted into a curated page.

Skip when:

• The task is general application testing unrelated to resource metadata.
• The task is schema authoring or code changes outside resource/.

Inputs

• Required: a specific resource file path such as resource/<id>/<id>.md.
• Optional: a related concern to emphasize, such as provenance, dates, products, or stub cleanup.

Workflow

1. Validate the file structurally.

   • Run:

     uv run make validate-file FILE=resource/<id>/<id>.md
     

   • If the file needs formatting cleanup, run:

     uv run make prettify-file FILE=resource/<id>/<id>.md
     uv run make validate-file FILE=resource/<id>/<id>.md
     

2. Check curation-specific quality rules that matter even when schema validation passes.

   • Confirm the file remains under resource/ and not registry/.
   • Confirm the page still has at least one products entry.
   • Confirm last_modified_date reflects today for the curation edit.
   • Confirm curators was not filled in.
   • If the page was formerly a stub, confirm domains: [stub], placeholder warnings, and stub prose were removed or replaced appropriately.

3. Check provenance fields carefully.

   • Inspect each product's original_source and secondary_source.
   • Each provenance entry must be a SourceAssociation object with both:
     • source
     • relation_type
   • source must be a KG-Registry resource ID or product ID, not a URL, INFORES ID, or descriptive label.
   • Use prov:hadPrimarySource as the default relation for original_source.
   • Use prov:wasInfluencedBy as the default relation for secondary_source when no more specific PROV-O relation is known.
   • Use prov:wasDerivedFrom when the product is transformed or repackaged from another source.
   • Use prov:wasInformedBy when another source influenced curation without being a direct derivation input.
   • Confirm directly owned products include the parent resource in original_source.
   • Confirm the parent resource is not redundantly listed in secondary_source for its own direct products.

4. Check referenced IDs for resolvability and intent.

   • When a provenance source references an existing KG-Registry resource, verify that the identifier matches the real registry ID.
   • When a provenance source references a not-yet-existing upstream resource, confirm the newly minted ID is unique, stable, and filename-safe.
   • A missing referenced ID can be acceptable during curation if it is an intentional new KG-Registry identifier. The tooling can later generate a stub page for it.
   • Do not silently replace a valid missing intended source ID with a raw URL or free-text label just to avoid creating a new identifier.

5. If validation fails, repair locally and rerun.

   • Fix the same file first.
   • Rerun uv run make validate-file FILE=... after each substantive fix.

Quick reference

Task	Command
Validate one file	uv run make validate-file FILE=resource/<id>/<id>.md
Prettify then validate	uv run make prettify-file FILE=resource/<id>/<id>.md && uv run make validate-file FILE=resource/<id>/<id>.md
Find provenance slots in a file	`rg -n 'original_source

Common mistakes

• Treating schema validation success as sufficient even when stub warnings or placeholder prose remain.
• Writing provenance as strings or labels instead of SourceAssociation objects.
• Using URLs or INFORES IDs in source instead of KG-Registry IDs.
• Forgetting that directly owned products should name their parent resource in original_source.
• Deleting a legitimate newly minted source ID because it does not yet have a page.

Relationship to other skills

• kg-registry-curation is the editing workflow.
• curate-next dispatches batches of curation tasks that should finish with this validation workflow.