AI Agent Skill¶

Uni ships an agent skill (uni-db) that gives AI coding assistants deep knowledge of the Uni API, Cypher dialect, Locy language, vector/hybrid search, Pydantic OGM, graph algorithms, and schema design. With the skill installed, your AI assistant can build Uni-powered applications without you having to paste documentation into every conversation.

The skill works with Claude Code, GitHub Copilot, Cursor, Cline, and other agents that support the skills.sh ecosystem.

What the skill provides¶

The skill uses progressive disclosure — a compact quick-reference is always loaded (~450 lines), and detailed references are loaded on demand only when relevant to the task:

Installation¶

From the skills.sh directory¶

npx skills add https://github.com/rustic-ai/uni-db --skill uni-db

This fetches the skill from the Uni repository and installs it into your agent's configuration. Works with any agent that supports the skills.sh directory.

Local installation (Claude Code)¶

If you have the Uni repository cloned locally, point your Claude Code settings to the skill directory:

{
  "skills": ["/path/to/uni-db/skills/uni-db"]
}

When the skill triggers¶

The skill activates automatically when the agent detects any of these signals:

• Code imports uni_db or uni_pydantic (Python) or depends on the uni-db crate (Rust)
• You mention "uni", "uni-db", or "embedded graph database" in context of code
• You write or ask about Cypher queries for a graph database
• You work with Locy programs or mention ALONG/FOLD/BEST BY/DERIVE/ASSUME/ABDUCE
• You need vector search, full-text search, or hybrid search on a graph
• You ask about graph algorithms like PageRank, shortest path, or community detection
• You ask about schema design, data types, indexes, or Pydantic OGM models

Usage examples¶

Once installed, just ask your agent what you need. The skill provides context automatically.

Build a schema:

Design a schema for a social network with users, posts, and follows.
Include vector embeddings on posts for semantic search.

Write queries:

Write a Cypher query that finds friends-of-friends who liked
the same posts, ordered by overlap count.

Set up hybrid search:

Add hybrid search (vector + full-text) over a Document label
with a 768-dim embedding and an abstract field. Use RRF fusion.

Write Locy rules:

Write Locy rules for transitive risk propagation through
a supply chain network using MNOR to combine probabilities.

Pydantic OGM models:

Using uni-pydantic, model a knowledge graph with Document nodes
that have vector embeddings and Category nodes connected by
TAGGED edges. Include lifecycle hooks and vector search.

Graph algorithms:

Run PageRank on my user network to find the most influential users,
then find communities using Louvain with a minimum size of 5.

Architecture¶

The skill is organized as a single SKILL.md entry point (~450 lines) with 8 reference files in a references/ directory:

skills/uni-db/
├── SKILL.md                      ← always loaded (routing + quick reference)
└── references/
    ├── cypher.md                 ← loaded for Cypher tasks
    ├── python-api.md             ← loaded for Python API tasks
    ├── rust-api.md               ← loaded for Rust API tasks
    ├── pydantic-ogm.md           ← loaded for Pydantic OGM tasks
    ├── vector-hybrid-search.md   ← loaded for search tasks
    ├── locy.md                   ← loaded for Locy tasks
    ├── schema-indexing.md        ← loaded for schema/index tasks
    └── graph-algorithms.md       ← loaded for algorithm tasks

The agent reads SKILL.md first, which contains a routing table that maps your task to the right reference file. Only the relevant reference is loaded into context, keeping token usage efficient.

Updating the skill¶

To get the latest version after a Uni release, re-run the install command:

npx skills add https://github.com/rustic-ai/uni-db --skill uni-db