Curate knowledge graph metadata following KG-Registry schema. Validate resources, expand stub entries, and maintain data quality standards.
This skill helps you curate knowledge graph metadata for the Knowledge Graph Registry (kg-registry). It guides you through expanding stub entries, validating schema compliance, and maintaining high-quality metadata about knowledge graphs and their relationships.
**ALWAYS follow these rules:**
1. **Use the canonical schema**: All Resource entries must follow `src/kg_registry/kg_registry_schema/schema/kg_registry_schema.yaml`
2. **Respect enums**: Only use values defined in the schema enums. Do not create new enum values.
3. **Use defined categories**: Do not create new categories. Use only the classes defined in the schema.
4. **Invalid categories**:
- ToolResource is NOT a valid category
- DataProduct is NOT a valid category (use "Product" if unsure)
5. **Never modify `registry/`**: This directory contains auto-generated content
**When creating or updating Resource entries:**
1. **Location**: Work only in the `resource/` directory
2. **Products field**: Always include at least one Product entry
3. **Date fields**:
- `creation_date`: Date the metadata page was created (today's date for new entries)
- `last_modified_date`: Today's date
- These dates refer to metadata creation, NOT the resource itself
4. **Curators field**: Leave empty (specific to KG-Registry team)
5. **Product pages**: Do NOT create individual Product pages manually - these are auto-generated from Resource entries
**When defining Products:**
1. **Original sources**: If a Product mentions another Resource (e.g., "A mapping file to WikiPathways"), find that Resource's ID and add it to the Product's `original_source` field
- Example: Product described as "mapping to WikiPathways" should have `wikipathways` in its `original_source` list
2. **Category assignment**: If unsure of category, use "Product" (generic fallback)
**Follow these steps to curate stub entries:**
#### Step 1: Identify Stub Resources
Find all Resource pages with `stub` in their `domains` list. These are incomplete entries awaiting curation.
#### Step 2: Select & Check
1. Select a stub Resource page at random
2. Check if it appears in `reports/curation_problems.tsv`
3. If listed in problems log, skip it and select another
#### Step 3: Research
1. Look for the `homepage_url` in the Resource page (good starting point)
2. Conduct literature search if needed
3. Gather comprehensive information about the Resource
#### Step 4: Expand the Entry
1. Consult the schema at `src/kg_registry/kg_registry_schema/schema/kg_registry_schema.yaml`
2. Use existing Resource pages in `resource/` as examples
3. Fill out all applicable fields
4. Include Products (list them, don't create Product pages)
5. Keep existing Products if present (may be referenced by other Resources)
6. **If you encounter blocking curation problems:**
- Log them in `reports/curation_problems.tsv`
- Move to another Resource
#### Step 5: Validate
1. Run validation: `poetry run make validate-file FILE=path/to/resource.yaml`
2. Fix all validity issues before proceeding
3. Do not commit invalid entries
#### Step 6: Repeat
Return to Step 2 and continue until you have curated at least **5 Resources**.
Use Poetry to run validation:
```bash
poetry run make validate-file FILE=resource/my_resource.yaml
poetry run make <target>
```
```yaml
id: example_kg
label: Example Knowledge Graph
description: A comprehensive knowledge graph for example data
domains:
- biomedical # Valid enum value from schema
homepage_url: https://example.org/kg
creation_date: "2026-02-03" # Metadata creation date (today)
last_modified_date: "2026-02-03" # Today's date
products:
- id: example_kg_nodes
label: Example KG Nodes
category: Product
description: Node list for Example KG
original_source:
- wikipathways # Another Resource this Product references
```
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/knowledge-graph-registry-curator-yhxglk/raw