What is Agent Knowledge?

Agent Knowledge is a portable directory format for packaging durable knowledge assets for AI agents.

It is designed for knowledge that should survive across sessions:

• brand and product facts
• organization know-how
• personal or expert profiles
• research wikis
• support and sales playbooks
• policy and compliance references
• long-lived domain context

It is not a replacement for Agent Skills. Agent Skills tell an agent how to perform work. Agent Knowledge tells an agent what facts, sources, context, and boundaries it may rely on.

The problem

Many systems put all knowledge into one of two places:

• a vector database with little human-readable structure
• a prompt or skill file that mixes facts with instructions

Both break down when knowledge must be maintained, reviewed, cited, and shared across agents.

Agent Knowledge separates layers:

raw sources -> maintained wiki -> compiled runtime views -> optional indexes

Core architecture

The Skill layer provides methods and workflows. The Knowledge layer provides source-grounded context. The agent runtime combines them only after trust, status, and grounding checks.

Core principles

1. Files first: a pack is a directory people and agents can inspect.
2. Sources stay separate: raw source material is evidence, not runtime prompt by default.
3. Knowledge is data: clients must treat loaded knowledge as context, not instructions.
4. Progressive disclosure: metadata first, usage guide second, context/evidence only as needed.
5. Indexes are rebuildable: vector, graph, and full-text indexes accelerate retrieval but are not facts.
6. Review state is explicit: draft, ready, stale, disputed, and archived knowledge behave differently.
7. Skills remain procedural: use skills to ingest, lint, query, and apply knowledge packs.