Home Directory as Knowledge Base

A structure for organizing your entire development environment around LLM-accessible domains and workflows.

The Structure

~/
├── repository             # Super-meta: Bootstrap scripts for new devices
├── ai_agents              # Orchestration layer (syncs commands to agents)
├── company_you_work_for   # Work-specific projects and context
├── auto-slacker           # Meta: LLM knowledge and this documentation
├── career-orchestrator    # Domain: Career/goal management workflows
├── design                 # Domain: Design patterns and decisions
├── developer              # Domain: Developer workflows and tools
├── operations             # Domain: Ops, deployment, infrastructure
├── research               # Domain: Research, experiments, learning
├── tester                 # Domain: Testing workflows and strategies
├── project_you_work_on    # Project: Specific application
├── workspace              # Config: VS Code workspaces and settings

The Pattern

Domain Repos

Organize by concern, not by project:

tester/ - Everything testing-related across all projects
developer/ - Development tools, workflows, practices
operations/ - Deployment, infrastructure, monitoring
design/ - Architecture decisions, patterns, documentation
research/ - Experiments, learning, investigations
career-orchestrator/ - Goals, career planning, meta-work

Each domain repo contains:

Commands that agents can execute
Documentation and patterns
Scripts and tooling
Examples and templates

Project Repos

Specific applications or work projects:

company_you_work_for/ - Work projects and company-specific context
project_you_work_on/ - Individual application
Any other standalone projects

Orchestration Layer

ai_agents/ - The meta-layer that:

References commands from domain repos
Syncs to agent-specific locations (~/.claude, ~/.cursor)
Coordinates across domains
Stays agent-agnostic

Super-Meta Layer

repository/ - The bootstrap layer:

Setup scripts for new devices
Configuration management
Device inventory
Meta scripts that manage everything

See: repository-meta-bootstrap.md for details.

Meta Layer

auto-slacker/ - The repository about repositories:

Documentation of the structure itself
LLM knowledge and patterns
Templates and examples
This file you’re reading

Why This Works

Knowledge Reuse: Testing patterns learned in one project immediately available in all projects

Context Sharing: LLMs can access domain knowledge regardless of current project

Single Source of Truth: Design decision in design/, not scattered across project READMEs

Composable Commands: Mix testing + deployment + design workflows across any project

Agent-Accessible: Everything structured for LLM consumption and reference

The Flow

Project Work
     ↓
Generates domain knowledge
     ↓
Stored in domain repos
     ↓
Referenced by ai_agents
     ↓
Synced to agent location
     ↓
Available in any project

Example Usage

Working on project_you_work_on:

Need to run tests → /coverage_check
Agent reads ~/.claude/commands/coverage_check.md
Follows pointer to ~/tester/agents/commands/coverage_check.md
Executes testing workflow from tester repo
Generates insights, stores in ~/tester/llm-lore/ or ~/research/
Knowledge now available for next project

The Meta Recursion

Working creates knowledge
Knowledge lives in domain repos
Domain repos feed commands
Commands improve workflows
Better workflows create better knowledge
Loop continues

The home directory organizes itself through use. The more you work, the better the structure becomes.

Practical Implementation

When starting new work:

Identify the domain - Is this testing? Design? Operations?
Use domain commands - Reference existing workflows from that domain repo
Capture new patterns - LLM writes discoveries back to domain repos
Update pointers - ai_agents references new commands
Sync propagates - Available immediately in agent

Let the LLMs maintain the domain repos. They understand the structure and can cross-reference.