rdf-generation - SKILL.md Agent Skill

name: rdf-generation description: Generate Turtle RDF from prose user input license: MIT compatibility: opencode metadata: workflow: semantic-web audience: developers focus: solid-project

What I do

Convert prose descriptions into valid Turtle (TTL) RDF syntax
Follow RDF best practices and semantic web standards
Use appropriate ontologies including Solid project vocabularies (ACL, VCARD, LDP, etc.)
Generate clean, well-formatted, and properly prefixed RDF
Validate turtle syntax before output
Suggest appropriate URIs and namespace conventions

When to use me

Use this skill when:

User asks to create RDF data
User wants to convert data descriptions to Turtle format
User needs semantic web/linked data representations
User requests ontology definitions or vocabulary files

Best practices I follow

1. Proper prefixes

Always declare common prefixes at the top. Include Solid project ontologies when relevant:

@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
@prefix dcterms: <http://purl.org/dc/terms/> .
@prefix foaf: <http://xmlns.com/foaf/0.1/> .
@prefix vcard: <http://www.w3.org/2006/vcard/ns#> .
@prefix acl: <http://www.w3.org/ns/auth/acl#> .
@prefix solid: <http://www.w3.org/ns/solid/terms#> .
@prefix ldp: <http://www.w3.org/ns/ldp#> .
@prefix schema: <http://schema.org/> .
@prefix as: <https://www.w3.org/ns/activitystreams#> .

2. URI conventions

Use meaningful, hierarchical URIs
Prefer HTTPS when possible
Prefer hash URIs (#) for vocabulary terms and resources
Use fragment identifiers for clean, bookmark-able resources
Example: https://example.org/vocab#Person rather than https://example.org/vocab/Person
For document-relative URIs, use <#fragment> or define a base URI with @base and use :fragment notation
When creating instance data, use hash URIs relative to the document (e.g., <#order1> or with @base, just #order1)

3. Literal datatypes

Always specify datatypes for typed literals
Common types: xsd:string, xsd:integer, xsd:date, xsd:boolean, xsd:decimal
Use language tags matching the user's input language (e.g., "Hello"@en, "Bonjour"@fr, "Hallo"@de)
Detect language from user's prose and apply appropriate tags
Use xsd:dateTime with ISO 8601 format

4. Blank nodes

Use blank nodes [] for complex nested structures
Prefer named nodes for resources that need to be referenced
Use blank node labels _:b1 only when necessary for readability

5. Turtle syntax

Use ; to continue with same subject
Use , to continue with same subject and predicate
Use . to end statements
Group related statements together
Add blank lines between different entities for readability

6. Common patterns

Classes:

:MyClass a rdfs:Class ;
    rdfs:label "My Class"@en ;
    rdfs:comment "Description of the class"@en ;
    rdfs:subClassOf :ParentClass .

Properties:

:myProperty a rdf:Property ;
    rdfs:label "my property"@en ;
    rdfs:comment "Description of the property"@en ;
    rdfs:domain :DomainClass ;
    rdfs:range :RangeClass .

Instances:

:instance1 a :MyClass ;
    rdfs:label "Instance 1"@en ;
    :myProperty :relatedInstance ;
    dcterms:created "2026-02-10T00:00:00Z"^^xsd:dateTime .

Questions to ask

When the user request is ambiguous, ask:

What is the base URI/namespace for this data?
Are there existing ontologies or vocabularies to reuse?
What is the intended use case for this RDF data?
Are there specific properties or relationships needed?
Should this follow specific Solid conventions (e.g., WebID profiles, ACL rules)?

Validation

Before finalizing output:

Check all prefixes are declared
Verify URIs are properly formatted
Ensure all statements end with .
Confirm datatypes are correct
Test that blank nodes don't create orphaned data

Output format

CRITICAL: Always create a .ttl file when generating RDF data.

Always output complete, valid Turtle files that can be:

Parsed by standard RDF libraries (Apache Jena, RDFLib, etc.)
Loaded into triple stores (Virtuoso, Blazegraph, GraphDB)
Validated using online validators or CLI tools

File creation requirements:

Save RDF data to a .ttl file with a meaningful filename
Use kebab-case for filenames (e.g., pizza-order.ttl, person-profile.ttl)
Place files in an appropriate location (ask user if unclear, default to current directory)

Example workflow

User: "Create RDF for a person named John Smith who works at Acme Corp"

Response: I'll create a Turtle file with RDF for John Smith.

@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix foaf: <http://xmlns.com/foaf/0.1/> .
@prefix vcard: <http://www.w3.org/2006/vcard/ns#> .

<#john-smith> a foaf:Person ;
    foaf:name "John Smith"@en ;
    vcard:organization-name "Acme Corp"@en .

Creates file: person-profile.ttl

Note:

Language tags match the user's input language. This example uses @en because the user wrote in English.
The resource uses a document-relative hash URI <#john-smith> instead of an external namespace prefix.