Turn messy domain knowledge into a staged ontology delivery workflow.

ontology_research is a hierarchical skill family for planning, researching, building, and validating domain ontologies with explicit review gates, saved evidence trails, and reusable execution artifacts.

This repo is for ontology work that is too consequential for a generic one-shot answer. It treats ontology engineering as an operational workflow: clarify the real decision context, gather evidence, model conservatively, validate before release, and preserve the trail in exe/.

Start with the root orchestration skill:

• ontology-agent-suite/SKILL.md

Use this first prompt if you want the full workflow:

Use ontology-agent-suite to set up a domain ontology workflow for a smart factory maintenance domain, including scope clarification, competency questions, evidence gathering, ontology construction, and validation checkpoints.

Read the methodology and shared rules when you want the full operating context:

• refs/Total Guide.md
• shared/references/skill-family-architecture.md
• shared/references/exe-artifact-contract.md

• Turn a vague domain problem into a scoped ontology plan with competency questions and review gates.
• Run structured source collection and evidence curation instead of jumping straight into schema design.
• Build a SKOS-first, reuse-first ontology with explicit modeling decisions before adding richer semantics.
• Validate the result with competency-question checks, SHACL-oriented constraints, and release-readiness review.
• Keep the working trail under exe/ so the process is inspectable, not just the final answer.

• ontology-agent-suite/SKILL.md: orchestrates the full ontology lifecycle across planning, research, build, and validation.

• ontology-planning-team/SKILL.md: runs the planning phase and coordinates scope, users, and review gates.
• ontology-domain-intake/SKILL.md: clarifies the target domain, scope boundaries, users, and constraints.
• ontology-cq-designer/SKILL.md: designs competency questions that later drive modeling and validation.
• ontology-technical-plan-writer/SKILL.md: turns planning outputs into a reusable technical ontology plan.

• ontology-research-team/SKILL.md: coordinates research, feasibility checks, reuse discovery, and evidence packaging.
• ontology-source-scout/SKILL.md: finds and triages standards, schemas, ontologies, and primary sources.
• ontology-evidence-curator/SKILL.md: converts collected material into ontology-ready evidence and modeling signals.

• ontology-build-team/SKILL.md: coordinates the transition from evidence to ontology draft.
• ontology-ontology-architect/SKILL.md: defines concepts, relations, mappings, and formalization strategy.

• ontology-validation-team/SKILL.md: checks release readiness across competency questions, structure, and governance.
• ontology-shacl-validator/SKILL.md: performs SHACL-oriented validation and reports structural issues.

• "Use ontology-agent-suite to design an ontology workflow for a healthcare claims domain with clear human review gates."
• "Use ontology-research-team to collect primary standards and ontology reuse candidates for an industrial maintenance knowledge model and save the trail under exe/."
• "Use ontology-build-team to turn curated evidence into a SKOS-first ontology outline with explicit modeling decisions."

• "Take this messy pile of standards, internal terms, and vendor schemas and turn it into a validation-ready ontology workflow with saved research traces and approval checkpoints."
• "Set up an ontology process for a regulation-heavy domain where mappings cannot be accepted until evidence and human review are both recorded."
• "Turn raw domain research into a reusable ontology delivery pack with competency questions, reuse candidates, modeling notes, and SHACL-oriented validation artifacts."

• It packages ontology engineering as a staged delivery workflow, not a loose set of prompts.
• It makes intent clarification and competency questions mandatory before artifact drafting.
• It preserves the research process in exe/, not only the polished answer.
• It separates planning, research, build, and validation responsibilities across different skills.
• It prefers stable conceptual modeling first, then formalization only when the evidence justifies it.
• It includes an evaluation surface for the orchestration skill instead of assuming the main prompt is already good enough.

1. Planning: clarify goals, users, scope boundaries, competency questions, and approval gates.
2. Research: collect sources, test data feasibility, log evidence, and identify reuse candidates.
3. Build: define concepts and relations, record modeling decisions, and formalize conservatively.
4. Validation: check competency-question coverage, SHACL-oriented structure, and release readiness.

This sequence is intentionally conservative. The repo is optimized for ontology work where traceability, reviewability, and stable modeling matter more than speed alone.

The root orchestration skill already has an evaluation scaffold:

• ontology-agent-suite/eval/profile.yaml
• ontology-agent-suite/eval/cases/
• ontology-agent-suite/eval/runs/run-20260331-002/

The current evaluation setup covers:

• happy-path end-to-end behavior
• regulated-domain review gating
• ambiguity and mapping-control behavior
• regression protection for core ontology principles

ontology_research/
├── refs/                           # synthesized methodology and source notes
├── exe/                            # run artifacts and saved working trail
├── shared/references/              # architecture, contracts, and templates
├── shared/scripts/                 # ontology and SHACL codegen helpers
├── ontology-agent-suite/           # root orchestration skill
├── ontology-planning-team/         # planning-stage coordinator
├── ontology-research-team/         # research-stage coordinator
├── ontology-build-team/            # ontology build-stage coordinator
├── ontology-validation-team/       # validation-stage coordinator
├── ontology-domain-intake/         # planning worker
├── ontology-cq-designer/           # competency-question worker
├── ontology-technical-plan-writer/ # planning document worker
├── ontology-source-scout/          # source discovery worker
├── ontology-evidence-curator/      # evidence curation worker
├── ontology-ontology-architect/    # ontology modeling worker
└── ontology-shacl-validator/       # SHACL-oriented validation worker

If you want to extend or adapt the skill family, the most important shared assets are:

• shared/references/templates.md
• shared/references/ontology-spec.template.json
• shared/references/shacl-spec.template.json
• shared/scripts/generate_ontology_ttl.py
• shared/scripts/generate_shacl_ttl.py

This repository currently includes:

• a root orchestration skill for end-to-end ontology workflow control
• stage-specific team and worker skills
• shared architecture and artifact contracts
• evaluation scaffolding for improving the orchestration layer

The biggest remaining proof asset is a polished, end-to-end example run that shows the full transformation from raw domain brief to validation-ready ontology artifacts.

If you are preparing this repository for public release, review:

• LICENSING.md