Library
The library is matic's immutable reference knowledge store — the layer where institutional knowledge solidifies into authoritative, reusable material. Unlike memory (mutable operational context an agent actively curates) or experience (an agent's private chronological work record), library entries are read-only during execution and change only through explicit PR-based curation. Libraries exist at org, project, and team scopes; agents consume from all scopes they belong to but write to none of them directly.
Library Schema
The complete frontmatter specification for library entries, including all required fields (id, scope, category, source, version) and optional fields that control retrieval ranking and grounding authority.
Entry Types
The five categories of library entries — reference, guide, standard, skill, and promoted-artifact — and when each is appropriate for a given piece of knowledge.
Filesystem Layout
How library directories are structured within org, project, and team scope handles, including the promoted-artifacts/ subtree and the conventions for domain, tools, standards, and skills subdirectories.
Indexing
How library entries are indexed at retrieval time by scanning frontmatter fields — no separate database; the filesystem and frontmatter are the index.
Retrieval Protocol
How agents query the library: during context envelope assembly before execution, and during grounding checks when validating factual claims in produced artifacts.
Promotion and Demotion
How project artifacts get promoted into the library via pull request and human review, and how existing library entries are replaced or versioned when they become stale.
Cross-Agent Sharing
How library visibility follows the scope hierarchy — org library is available to all agents, project library to project members, team library to team members — and how this differs from memory inheritance.