OntoStore

The official skill registry for OntoSkills.

Browse, install, and publish validated OWL 2 skill packages for deterministic AI agent behavior.

Documentation • Browse Skills • Publish

---

What is OntoStore?

OntoStore is the versioned skill registry at the heart of the OntoSkills ecosystem. It hosts pre-compiled OWL 2 ontology packages that AI agents query via SPARQL for deterministic, verifiable behavior.

Every package contains .ttl artifacts validated against the OntoSkills SHACL constitution — no probabilistic interpretation, no ambiguity.

ontoskills install obra/superpowers          # Install a package
ontoskills install obra/superpowers --with-embeddings  # With semantic search files
ontoskills install anthropics                # Install all from an author

---

Available Packages

Author	Package	Skills	Category
anthropics	claude-office-skills	137	Office & productivity
anthropics	knowledge-work-plugins	113	Knowledge management
anthropics	financial-services-plugin	27	Finance & compliance
anthropics	claude-plugin-official	19	Plugin development
anthropics	claude-code	10	Code assistance
coreyhaines31	marketingskills	34	Marketing
pbakaus	impeccable	21	Quality & testing
coinbase	agentic-wallet-skills	9	Blockchain & wallets
n8n-io	n8n	10	Workflow automation
obra	superpowers	14	Developer tooling
nextlevel-builder	ui-ux-pro-max-skill	7	UI/UX design
kotlin	kotlin-agent-skills	4	Kotlin development
remotion-dev	skills	1	Video creation
vercel-labs	vercel-react-best-practices	1	React development
vercel-labs	agent-browser	1	Browser automation
mareasw	greeting	1	Demo / hello world

---

Trust Tiers

Every package declares a trust tier that affects search ranking and visibility:

Tier	Multiplier	Description
official	1.2x	Maintained by the OntoSkills team
verified	1.0x	Reviewed and validated from known vendors
community	0.8x	Community-contributed, use at your own discretion

---

Skill Discovery

Installed skills are discoverable by the OntoMCP server through two search engines:

BM25 Keyword Search (default, always available)

Every installed skill is searchable via BM25 keyword matching — no extra dependencies, no model downloads. The BM25 index is built in-memory at server startup from skill intents, aliases, and descriptions.

• Zero setup — works out of the box after ontoskills install
• English stemming and stop words
• Results ranked by keyword relevance x trust-tier quality multiplier

Semantic Search (optional)

For large skill catalogs where keyword matching may miss relevant results, OntoStore provides pre-computed embeddings. Install them with --with-embeddings:

ontoskills install obra/superpowers --with-embeddings

When available, the MCP server uses ONNX inference for semantic similarity matching. BM25 remains the default; semantic search activates only when embeddings are installed.

---

Registry Structure

ontostore/
├── index.json                         # Registry index (packages list + embedding model)
├── embeddings/
│   └── tokenizer.json                 # Shared tokenizer (for semantic search)
├── packages/
│   ├── anthropics/
│   │   ├── claude-code/
│   │   │   ├── package.json           # Package manifest
│   │   │   └── <skill>/
│   │   │       ├── ontoskill.ttl      # Compiled OWL 2 ontology
│   │   │       └── intents.json       # Pre-computed embeddings (optional)
│   │   └── ...
│   ├── obra/
│   │   └── superpowers/
│   │       └── ...
│   └── ...
└── README.md

index.json

The registry index lists all available packages and the embedding model metadata:

{
  "embedding_model": {
    "model_name": "sentence-transformers/all-MiniLM-L6-v2",
    "dimension": 384,
    "model_file": "model.onnx",
    "tokenizer_file": "tokenizer.json"
  },
  "packages": [
    {
      "package_id": "obra/superpowers",
      "trust_tier": "verified",
      "source_kind": "ontology",
      "manifest_path": "packages/obra/superpowers/package.json"
    }
  ]
}

package.json (per-package manifest)

Each package has a manifest declaring its skills, modules, and metadata:

{
  "package_id": "obra/superpowers",
  "version": "0.11.0",
  "trust_tier": "verified",
  "source_kind": "ontology",
  "modules": ["tdd/ontoskill.ttl", "brainstorming/ontoskill.ttl"],
  "skills": [
    {
      "id": "tdd",
      "path": "tdd/ontoskill.ttl",
      "default_enabled": true,
      "aliases": ["test-driven-development"],
      "intents": ["write tests before code", "practice TDD"]
    }
  ],
  "embedding_files": ["tdd/intents.json", "brainstorming/intents.json"]
}

---

Installing Packages

From the CLI

# Install a single package
ontoskills install obra/superpowers

# Install with semantic search embeddings (optional)
ontoskills install obra/superpowers --with-embeddings

# Install all packages from an author
ontoskills install anthropics

# Enable/disable specific skills
ontoskills enable obra/superpowers tdd brainstorming
ontoskills disable obra/superpowers tdd

Adding a Registry Source

ontoskills registry add-source official https://ontoskills.sh/ontostore/index.json

---

Publishing a Package

1. Compile your skills

ontocore compile -i skills/ -o ontoskills/

2. Create a package.json

Place it in your skill directory alongside the compiled .ttl files. See the package spec for the full schema.

3. Submit a PR

Fork this repository, add your package under packages/<author>/<package>/, update index.json, and open a pull request. Packages are validated against the OntoSkills SHACL constitution before merge.

Requirements:

• All .ttl files must pass SHACL validation
• package.json must conform to the package manifest schema
• Skill IDs must be unique across the registry
• Trust tier will be set to community by default; verified requires review

---

Related

• OntoSkills — The platform
• OntoCore — The compiler
• OntoMCP — The MCP runtime
• CLI Reference — Install commands

---

© 2026 MareaSW