# 1000xagent — Master Knowledge Base

**Last Updated:** April 16, 2026 (TODO #20 — first-person pass, folder-map reference added)

---

## Overview

For detailed documentation:

- **Who I am:** `~/.claude/1000xagent/IDENTITY.md`
- **Agent inventory:** `~/.claude/1000xagent/agents.yaml` (single source of truth for all agents, businesses, categories, site display data)
- **Terminology:** `~/.claude/1000xagent/TERMINOLOGY.md` (glossary of the load-bearing terms I use)
- **My folder structure:** `~/.claude/1000xagent/folder-map.md` (classifies every path in `~/.claude/` as mine / utility / reference)
- **Deprecated patterns:** `~/.claude/1000xagent/deprecated-patterns.md` (things I've retired, so I don't reinvent them)
- **Rules:** `~/.claude/rules/` (auto-loaded principles and architecture)
- **State:** `~/.claude/state/latest.json` (session persistence via /save-state)
- **Memory:** `~/.claude/projects/-home-kxdev/memory/` (persistent cross-session memory)

---

## Tyler's Context

- **Name:** Tyler Youk
- **Position:** Founder & CTO of the 1000x ecosystem
- **Location:** Northern Virginia
- **Vision:** 1000x developer through AI leverage — he has created me to help him achieve 1000x output across software engineering and the rest of his life

---

## System Environment

- **WSL2 on Windows** — my primary runtime
- PowerShell available: `powershell.exe -Command "..."`
- Java 21 installed (for Firebase emulators)
- Node.js + npm in WSL
- `claude` CLI aliased to `claude --dangerously-skip-permissions` in `~/.bashrc` (bypass-permissions policy; see `~/.claude/rules/permission-policy.md`)

---

## My Folder (Key Locations)

```
~/.claude/                     # This folder is mine
├── CLAUDE.md                  # Entry point (loaded first)
├── rules/                     # My behavioral rules (auto-loaded)
├── skills/                    # My cross-cutting skills (descriptions auto-loaded)
├── commands/                  # My own commands + cross-cutting utilities
│                              #   NEVER project-agent activation commands
├── agents/                    # Only 1000xagent.md lives here
├── private/                   # Secrets (gitignored) — Vercel token, etc.
├── state/                     # My session state (latest.json)
├── research/                  # My research artifacts
├── context/                   # Additional context I reference
├── hooks/                     # My session hooks
├── logs/                      # My session logs
├── projects/                  # Auto-memory across my sessions
└── 1000xagent/                # My knowledge base
    ├── IDENTITY.md            # Core identity
    ├── KNOWLEDGE.md           # This file
    ├── TERMINOLOGY.md         # Glossary
    ├── folder-map.md          # Structural map
    ├── deprecated-patterns.md # Retired patterns
    └── agents.yaml            # Agent registry
```

For the full classification of every file and folder here, read `folder-map.md`.

---

## Skills vs Commands

- **New recurring workflow → Skill** (auto-loaded, model-invocable)
- **System operation or dashboard → Command** (user-invoked only)
- See `~/.claude/rules/skill-management.md` for full framework
- Project agents have their OWN skills and commands in their own repos — I do not host them

---

## Operating Patterns

### Notes Processing

Tyler writes tasks in `~/TODO.md`. I process them with `/spec` → `/solve` (auto-commits and removes the TODO on completion).

### State Management

- `/save-state` writes my session state to `state/latest.json`
- `/load-state` reads it and presents a resumption dashboard
- Memory system in `projects/-home-kxdev/memory/` persists facts across conversations (user, feedback, project, reference types)

### Git Operations

- I have full push authority across all repos
- `/1000xagent:push-all` pushes every dirty repo under `~/dev/`
- My commits in my own `~/.claude/` use `KX Developer`; commits in project repos use `tyleryouk <[redacted-email]>` (Vercel Hobby convention; see memory `feedback_vercel_git_author.md`)
- No `Co-Authored-By` on deploy-path commits (Vercel rejects)

### Deployment

- Vercel CLI with token at `~/.claude/private/vercel-token.txt`
- `/vercel-deploy` skill handles build + deploy + domain setup
- Scope: `[your-vercel-team]`

### Invocation Boundary

- From `~/`: I am the only agent available. Tyler invokes `/1000xagent` and my `/1000xagent:*` commands, plus the cross-cutting utilities.
- From any project repo under `~/dev/{repo}/`: that repo's project agent is active. I am still accessible via the utility commands (/spec, /solve, etc.) but the project agent is the primary context.