Ontology Skills

ontology-building

A domain-agnostic methodology, packaged as an [Agent Skill](https://code.claude.com/docs/en/skills), to assist in **building, extending, reviewing, and refactoring** ontologies, domain models, taxonomies, and knowledge-graph schemas.

It is deliberately independent of what you encode the result in, whether OWL, a property graph, a relational schema, typed code, or a platform object model. The approach encourages a focus on *what should exist and why*; the encoding comes after.

Why this exists

Guidance on ontology design is scattered across lineages that rarely talk to each other: the classical knowledge-representation literature, formal ontology theory, enterprise conceptual modelling, large-scale multi-team ontology governance, and — more recently — the practice of building models that AI agents read and act on. They agree more than they disagree, but where they disagree, most write-ups quietly pick a side.

The work of ontology building is still essentially 'human', but the universe of those with experience remains small, so tools that help more people engage in the process of capturing domains with logical structure help build the rails for AI.

This skill synthesises multiple facets of the approach into a single process and, where they genuinely conflict, says so explicitly and gives a precedence rule. `references/conflicts-and-precedence.md` documents eleven such conflicts and their resolutions.

What's in it

Four **operating modes** — Build, Extend, Review/audit, and Refactor — over a common seven-step process, plus:

• Competency-question scoping, with schema/system requirements treated as a peer input rather than an afterthought
• Mechanical tests for the hard calls: is-a litmus, subclass-vs-property, decomposition, taxonomy-artifact, reify-or-not
• OntoClean grounding (identity, rigidity, unity, dependence) for hierarchy disputes
• Definition-writing discipline — genus-differentia, necessary vs. sufficient condi