ontocore 1.0.0

OntoCore — Neuro-symbolic skill compiler. Compiles SKILL.md into validated OWL 2 ontologies.

OntoCore

Deterministic AI Skills via OWL 2 Ontologies

🇬🇧 English • 🇨🇳 中文

The neuro-symbolic compiler for the OntoSkills platform.

Compile natural language skills into verified, queryable knowledge graphs —
an alternative to probabilistic agent skills with a lightning-fast Rust MCP.

---

What is OntoCore?

OntoCore is the Python compiler at the heart of the OntoSkills platform. It acts as a neuro-symbolic compiler that transforms unstructured, human-readable AI skills (SKILL.md) into strictly validated, queryable OWL 2 ontologies.

By combining the natural language understanding of LLMs with the deterministic formal logic of RDF and SHACL validation, OntoCore ensures that AI agents operate on exact, verifiable knowledge graphs rather than probabilistic prompts.

Key Capabilities

• LLM Knowledge Extraction: Extracts structured triples (Dependencies, Inputs, Intents, Operations) from markdown files.
• SHACL Validation: Ensures the extracted semantic graph strictly adheres to the OntoSkills Core Ontology.
• Structural Content Preservation: Extracts code examples, tables, flowcharts, templates, and ordered procedures from markdown — losslessly, via deterministic parsing.
• OWL 2 Compilation: Outputs self-contained .ttl (Turtle) graphs ready for deterministic SPARQL querying.
• Local Registry Management: Handles the installation, enabling, and indexing of distributed skills packages.
• Security Auditing: Analyzes the graph for conflicting intents, missing dependencies, or shadowed skills.

---

Installation

Install the compiler directly from PyPI (requires Python 3.10+):

pip install ontocore

---

Quick Start

1. Initialize the Environment

Create the necessary folder structure (.ontoskills/) in your project:

ontocore init-core

2. Configure the LLM

OntoCore needs an LLM to extract relationships. Create a .env file or export the keys:

export OPENAI_API_KEY="sk-..."

(Anthropic is also supported via ANTHROPIC_API_KEY)

3. Compile Skills

Assuming you have SKILL.md files in a skills/ directory, run the compiler:

ontocore compile

This will read the markdown files, extract knowledge, validate it via SHACL, and generate .ttl ontology files in the .ontoskills/ output directory.

4. Query the Knowledge Graph

You can perform exact graph queries using SPARQL directly from the CLI:

ontocore query "SELECT ?skill WHERE { ?skill oc:resolvesIntent 'create_pdf' }"

---

CLI Reference

The package provides the ontocore command-line tool. Here are the main commands:

Core Commands

• ontocore compile: Compile local skills to validated OWL 2 ontologies.
• ontocore query <sparql_query>: Execute a SPARQL query against the compiled domain graph.
• ontocore security-audit: Run security checks against the knowledge graph to find issues.
• ontocore init-core: Initialize an empty OntoSkills registry in the current directory.
• ontocore list-skills: List all successfully compiled skills in the domain graph.

Registry & Packages

• ontocore install-package <path>: Install a .tar.gz skill package.
• ontocore import-source-repo <url>: Import skills directly from a remote Git repository.
• ontocore install: Download and install all dependencies declared in the lockfile.
• ontocore enable <skill_id>: Enable an installed skill.
• ontocore disable <skill_id>: Disable an installed skill.
• ontocore list-installed: Show all installed packages and their status.
• ontocore rebuild-index: Rebuild the registry index manually.

Run ontocore --help or ontocore <command> --help for detailed usage.

---

Documentation & Source

• OntoCore docs — Architecture, compilation pipeline, and SHACL validation
• Getting Started — Full installation and first steps
• mareasw/ontoskills — Source code and contributions

---

© 2026 Marea Software