Ontology Creator Skill for Claude & AI Agents


The Ontology Creator skill (skill-ontology-creator) generates ontological knowledge graphs in [[wikilinks]] syntax — entities, classes, relationships, axioms, and rules — either for a topic you name or extracted from a text you provide. The output is formatted for direct paste into InfraNodus, where it becomes a network visualization you can develop further with AI-powered gap analysis.

It's for knowledge engineers, researchers, note-takers, and anyone building structured knowledge bases who wants ontologies that look like actual networks — not the hub-and-spoke trees LLMs produce by default.

View Skill on GitHub Download skill-ontology-creator.zip InfraNodus MCP Server

What the Skill Does

Once installed, prompts like these trigger the skill:

  • "Create an ontology of climate change" — generates a comprehensive ontology for the domain, covering the widest possible range of entities and relation types.
  • "Extract entities and relationships from this text" — mines the explicit and implicit relations from any text you paste.
  • "Expand the renewable energy part" — follow-up requests develop the ontology in the direction you name, in the same syntax.
  • "Visualize this as a knowledge graph" — hands the ontology to the InfraNodus MCP tools for graph generation and metrics (if connected).
  • "Save this ontology to my InfraNodus account / as a memory" — persists the graph or stores the relations as structured memory for later retrieval.

The output is the ontology and nothing else — no explanations, no commentary — in a code snippet you can copy straight into InfraNodus.

How It Works

Every statement in the output follows one pattern: two or more entities in [[wikilinks]], a plain-text relation description, and a relation code in square brackets at the end of the line:

[[apple]] is an instance of [[fruit]] [isA]
[[apple]] grows as a result of [[apple blossom]] [causedBy]
[[chlorophyll]] absorbs [[sunlight]] for energy [relatedTo]

The skill uses a fixed vocabulary of ten relation codes — [isA], [partOf], [hasAttribute], [relatedTo], [dependentOn], [causes], [locatedIn], [occursAt], [derivedFrom], [opposes] — unless you provide your own. Each relation type gets multiple statements (8+ per type), balanced between descriptive aspects (classes, attributes, locations) and functional ones (causal chains, dependencies, rules).

The key design principle is network structure over trees. Instead of radiating every statement from one central concept (a star topology where "machine learning" appears in every line), the skill distributes entities across interconnected relationship chains: A relates to B, B affects C, C occurs during D. That's what makes the resulting graph useful in InfraNodus — topical clusters and structural gaps only emerge when the ontology is a genuine network.

The skill works standalone: paste the output into InfraNodus.com and the [[wikilinks]] become nodes, the co-occurrences become edges. If the InfraNodus MCP server is connected, the skill can skip the copy-paste: it uses generate_knowledge_graph for a one-off analysis (whose metrics it can also use to rebalance the ontology), create_knowledge_graph to save the graph to your account, memory_add_relations / memory_get_relations to store ontologies as structured memory, and search to connect a new ontology to graphs you already have.

Install the Skill

Claude Code

Download the skill into your global skills folder:

cd ~/.claude/skills
curl -L -o skill-ontology-creator.zip https://github.com/infranodus/skills/releases/latest/download/skill-ontology-creator.zip
unzip skill-ontology-creator.zip -d skill-ontology-creator

Then ask Claude to "create an ontology" of any topic or text.

Claude Web / Desktop

Go to Settings > Capabilities, enable Code execution and file creation, scroll to the Skills section, and upload the skill-ontology-creator.zip file.

OpenClaw

Run in the OpenClaw chat:

install this skill: https://github.com/infranodus/skills/releases/latest/download/skill-ontology-creator.zip

See the main installation guide for step-by-step instructions per LLM client.