Laraskills

Glossary

Key terms and concepts used throughout the Laraskills documentation and engineering knowledge system.

Operating Layer

The top-level integration layer that connects AI coding agents to Laravel engineering knowledge. It defines how agents discover, load, and apply skills, rules, and context during development sessions.

Intelligence Layer

The structured body of curated Laravel engineering knowledge, including knowledge units, rules, skill workflows, and decision trees. This is the content that Laraskills installs and makes available to AI agents.

Retrieval Layer

The deterministic search and retrieval system that finds the most relevant knowledge units, rules, and skills for a given task description. Runs entirely locally with no external API calls.

MCP

Model Context Protocol — an open protocol that allows AI coding agents to communicate with tools and data sources. Laraskills implements a read-only MCP server for deterministic local knowledge retrieval.

Knowledge Unit

A self-contained piece of Laravel engineering knowledge with a canonical ID, content body, metadata, prerequisites, and related topics. Knowledge units are the atomic building blocks of the intelligence layer.

Routing Map

A graph structure that defines relationships between knowledge units — including prerequisites and related topics. The routing map enables contextual navigation through the intelligence layer.

Skill Workflow

A structured, repeatable capability that bundles knowledge units, rules, and retrieval workflows. Skill workflows give AI agents the context they need to perform specific Laravel development tasks.

Required-Context Mode

A retrieval mode where the system enforces that all prerequisite knowledge units are included in the context bundle before the requested topic, ensuring the AI agent has complete background context.

Deterministic Retrieval

A retrieval approach where the same input always produces the same output. Unlike AI-powered semantic search, deterministic retrieval uses explicit graph traversal and rule-based matching for reproducible results.