From kg-registry-curation
Curates a single KG-Registry resource page from stub or incomplete state to complete metadata, following LinkML schema constraints.
How this skill is triggered — by the user, by Claude, or both
Slash command
/kg-registry-curation:kg-registry-curationThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Curate a single KG-Registry resource page end to end.
Curate a single KG-Registry resource page end to end.
resource/, consulting local repository context, and running local validation commands.resource/.Skip when:
registry/.resource/<id>/<id>.md.If the input is ambiguous, resolve it before editing. Do not guess between multiple candidate resources.
Resolve the target file.
resource/.Confirm the governing constraints before editing.
src/kg_registry/kg_registry_schema/schema/kg_registry_schema.yaml.registry/; those files are generated.products entry.curators field.creation_date as the date the KG-Registry metadata page was created.last_modified_date to today when making updates.Gather only the minimum evidence needed for this resource.
homepage_url, repository, existing products, publications, and warnings.resource/ as structural examples.Expand or correct the YAML front matter.
.yaml.layout: resource_detail unless the file already uses a different valid resource layout.category.description, homepage_url, repository, license, contacts, publications, domains, tags, taxon, and products when supported by evidence.infores_id when you can verify the exact identifier in the authoritative INFORES catalog at https://biolink.github.io/information-resource-registry/infores_catalog.yaml or its checked-in local cache. Do not infer or guess an INFORES ID from the KG-Registry resource ID, resource name, homepage, or ontology prefix.infores_id unset rather than minting a plausible-looking value.original_source and secondary_source as lists of SourceAssociation objects with both source and relation_type.source, not homepage URLs, INFORES IDs, free text labels, or citations.original_source is for the direct primary source or sources of the product. The default relation type is prov:hadPrimarySource.secondary_source is for additional upstream influences, aggregators, or contextual sources that affected the product without being the primary source. The default relation type is prov:wasInfluencedBy.prov:wasDerivedFrom fits transformed or re-packaged products and prov:wasInformedBy fits informative but indirect inputs.original_source for products that are owned directly by that resource.secondary_source when it is the direct owner of the product.Product.Update the Markdown body only as needed.
Prettify and validate the file.
kg-registry-validation skill for validation and quality checks.uv run make prettify-file FILE=<target-file> after editing when formatting needs normalization.uv run make validate-file FILE=<target-file> before finishing.If curation is blocked, handle it explicitly.
reports/curation_problems.tsv.resource_id, reason, details, and date_attempted.no_homepage, insufficient_info, or page_unresolvable.| Task | Command |
|---|---|
| Validate one file | uv run make validate-file FILE=resource/<id>/<id>.md |
| Prettify one file | uv run make prettify-file FILE=resource/<id>/<id>.md |
| Find stub files | rg -l '^\s*-\s*stub\s*$' resource |
| Check prior curation failures | cut -f1 reports/curation_problems.tsv |
registry/ outputs directly instead of the source file under resource/.original_source or secondary_source instead of KG-Registry identifiers.infores_id that was inferred from the local KG-Registry ID or ontology prefix without confirming the exact value in https://biolink.github.io/information-resource-registry/infores_catalog.yaml.secondary_source where original_source is appropriate, or omitting the parent resource from a directly owned product's primary provenance.curators even though that field is reserved for the KG-Registry team.creation_date or last_modified_date.domains: [stub] in place after a successful curation.uv run make validate-file FILE=... before finishing.curate-next selects the next stub resources and dispatches this single-resource workflow.kg-registry-validation covers per-file validation, provenance checks, and quality-review steps after editing.kg-registry-curation is the core per-resource curation procedure.npx claudepluginhub knowledge-graph-hub/kg-registry --plugin kg-registry-discoveryValidates KG-Registry resource pages for schema compliance, provenance slot correctness, and stub-cleanup issues after curation.
Maintains Obsidian-based LLM-driven wiki: ingests research papers/sources, compiles knowledge, manages topics/milestones/cross-references, queries wiki, runs lint checks.
Processes claimed captures into karpathy-wiki pages via spawned Claude ingester. Runs 9-step orientation: reads schema/index/log, extracts signals from capture, scores/rank/read candidates, decides create/augment/no-op.