AI

Enterprise-grade reference materials for AI-assisted development teams.

This repository is your onboarding bridge, command center, and guardrail layer for building and running AI agent workflows with consistent quality and speed.

What you get

• Guidance-first operations: AGENTS + workspace templates keep every directory aligned.
• Reliable bootstrap: ai init wires in shared conventions in one step.
• Fleet-wide updates: Registry tooling updates every initialized directory together.
• Production posture: Explicit architecture, testing, and review expectations baked in.

Start here (recommended)

1. Install the CLI
2. Initialize your project directory
3. Run updates as your agent network evolves

1) Install

Install with npx (no global dependencies required):

npx @sethwebster/ai-cli

Or install globally:

npm install -g @sethwebster/ai-cli

Then reload your shell and initialize:

source ~/.zshrc  # or source ~/.bashrc
ai init

2) Alternative install paths

If you prefer script-based setup:

curl -fsSL https://raw.githubusercontent.com/sethwebster/AI/main/install.sh | bash

Or install only the shell function manually:

curl -fsSL https://raw.githubusercontent.com/sethwebster/AI/main/ai-function.sh >> ~/.zshrc
source ~/.zshrc
ai init

Project structure

AI/
├── agents/              # Agent definitions
├── bin/
│   └── install.js       # npm installer script
├── landing-page/        # Documentation website
├── migrations/          # Migration system
├── AGENTS.md            # Universal development guardrails
├── AGENT-WORKSPACE.md   # Workspace-specific template
├── ai-function.sh       # Shell function definition
├── install.sh           # Installer script
├── package.json         # npm package config
└── README.md            # This file

CLI command map

ai init        # Prepare current directory with AGENTS + workspace template
ai update      # Pull latest AI repo changes into local clone
ai reload      # Reload shell profile so ai function updates take effect immediately
ai update-all  # Update all directories registered for management
ai list        # Show all registered directories
ai forget      # Remove current directory from registry

What ai init actually does

1. Updates ~/.ai-repo-local-clone (shallow clone) to keep references current.
2. Creates stable symlinks in the target directory:
   • AGENTS.md → ~/.ai-repo-local-clone/AGENTS.md
   • CLAUDE.md → ~/.ai-repo-local-clone/CLAUDE.md
3. Copies AGENT-WORKSPACE.md from template (only when missing).
4. Registers the directory in ~/.ai-registry for bulk operations.

That is why ai update remains your safest path for synced guidance updates.

Registry model

The registry is a plain-text file at ~/.ai-registry with one directory per line.

• ai list
  Enumerates active managed directories and marks missing paths.
• ai reload
  Reloads your shell profile (~/.zshrc/~/.bashrc) so local ai function updates apply right away.
• ai update-all
  Applies updates across all tracked directories in one run.
• ai forget
  Removes the current directory from the registry without deleting local files.

Who should read what first

For AI agents

Read in this order:

1. AGENTS.md – universal safety, quality, and architecture expectations.
2. AGENT-WORKSPACE.md – project-specific patterns and current context.

For developers

Use AGENTS.md when you are:

• Designing new AI agent systems.
• Making architecture decisions (ADRs included).
• Defining quality standards or review policy.
• Onboarding teams to production-style AI workflows.

Then use AGENT-WORKSPACE.md to capture:

• Local setup and command conventions.
• Active work tracking.
• Project-specific operating rules.

Key principles to keep in view

• Explicit over implicit.
• Fail fast, fail loud.
• Optimize for deletion.
• Test-first development mindset.
• ADR discipline for architectural decisions.
• Strict typing and disciplined error boundaries.

Why this matters

Without a stable baseline, AI projects drift into inconsistent quality, hidden assumptions, and avoidable rework.
This repository keeps teams moving in the same direction: faster iteration, better decision records, and fewer downstream surprises.