lr

Lore Framework

A persistent knowledge system for AI agents running in Claude Code.

Agents accumulate lore — domain expertise, operational wisdom, and decision history — across sessions. Each agent's knowledge graph lives in a shared git repository, growing through contributions from every teammate who boots the agent.

Team-Shared Knowledge

Lore agents are team-shared knowledge containers, not personal notebooks. The framework's purpose is to convert tribal knowledge — domain expertise, design rationale, decisions, status, intermediate work — into durable, transmissible assets stored in shared git repos.

Multiple contributors are expected to:

• Boot the same agent in their own sessions
• Reflect and merge into the same lore/ and lore-context.md
• Commit sessions/ summaries that teammates will read for context
• Push concurrently — conflicts are auto-resolved via the merge process

This framing informs every design choice: directory-driven storage so git is the medium, plain markdown so anyone can read and edit, push-conflict-resolution because concurrent contributors are expected (not exceptional), sessions as narrative artifacts written for future readers.

How It Works

The framework provides:

• A structured way to define AI agents with persistent memory
• A reflection/merge process that extracts and integrates session knowledge
• Skills for managing agents and their knowledge lifecycle

Agents are plain markdown. Knowledge is plain markdown. Git tracks everything. No databases, no config files, no build steps.

Concepts

A handful of terms appear throughout the docs:

• Workspace — the directory you run Claude Code from. Holds one or more agent repos and any other repos they declare.
• Agent repo — a git repo containing one or more lore agents. Marked by a lore-repo.md at its root. Conceptually, the domain an agent (or set of agents) covers.
• Agent — a directory under agents/<name>/ inside an agent repo. Has a role.md (identity) and lore-context.md (working knowledge), plus a lore/ knowledge graph of markdown topics.
• Lore — the agent's accumulated knowledge: decisions, domain expertise, operational wisdom. Plain markdown, tracked in git, shared across teammates.
• Boot — load an agent's role and lore into your Claude Code session via /lr:boot <name>.
• Finalize — at session end, extract what was learned, merge it into the agent's lore, write a session summary, and commit + push. One command: /lr:finalize.

Installation

Install the lr plugin in Claude Code:

# Add the lore-framework marketplace
claude plugin marketplace add zroslaw/lore-framework

# Install the lr plugin
claude plugin install lr@lore-framework

For local development:

claude --plugin-dir ./lore-framework

Quick Start

Pick the path that matches your situation.

Joining a team that already uses Lore Framework

A teammate has set up an agent repo and pointed you at it.

1. Clone any one of the agent repos into a workspace directory of your choice.
2. Run Claude Code from the workspace (the parent directory).
3. Sync the workspace — clones any other repos the agent repo declares; pulls everything:

   /lr:workspace-sync
   

4. Initialize the workspace so future Claude Code sessions auto-load the framework's conventions:

   /lr:init
   

5. Boot an agent and start working:

   /lr:boot <agent-name>
   

   (Run /lr:list-agents if you don't yet know what's available.)
6. Finalize at session end to preserve what was learned:

   /lr:finalize
   

Starting fresh — creating your own agent repo

You're introducing the framework into a new area.

1. Run Claude Code from a workspace directory.
2. Create an agent repo:

   /lr:create-repo my-agents
   

3. Create an agent:

   /lr:create-agent
   

4. Initialize the workspace:

   /lr:init
   

5. Boot and work with the agent:

   /lr:boot my-agent-name
   

6. Finalize at session end:

   /lr:finalize
   

Skills

Grouped by purpose:

Workspace setup

Skill	Purpose
/lr:workspace-sync	Clone declared repos and pull all top-level repos in the workspace
/lr:init	Write the framework-managed section into the workspace's CLAUDE.md
/lr:list-agents	List all agents in the workspace
/lr:list-repos	List all agent repos in the workspace

Working with agents

Skill	Purpose
/lr:boot <name>	Load a lore agent by name
/lr:recall [hint]	Search lore across loaded agents
/lr:attach <agent>	Attach another agent as a guest in this session
/lr:consult <agent> [hint]	One-shot question to an unloaded agent
/lr:spawn-teammate [<agent>...]	BETA — Spawn lore agents as Agent Teams teammates