Skip to content

pai-core-install

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History
 
 

pai-core-install

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
src/skills/CORE
src/skills/CORE
 
 
INSTALL.md
INSTALL.md
 
 
README.md
README.md
 
 
VERIFY.md
VERIFY.md
 
 

README.md

name PAI Core Install
pack-id danielmiessler-pai-core-install-v2.3.0
version 2.3.0
author danielmiessler
description Complete CORE skill installation - the foundational skill that governs PAI system operation, architecture, and all system-level configuration
type skill
purpose-type
infrastructure
configuration
foundation
platform claude-code
dependencies
keywords
core
foundation
system
architecture
configuration
response-format
workflows
delegation
skills
memory
hooks
security
pai

PAI Core Install

PAI Core Install

The foundational skill that governs PAI system operation - auto-loads at session start to provide architecture documentation, response formats, workflow routing, and system-level configuration.

Installation Prompt

You are receiving a PAI Pack - a modular upgrade for AI agent systems.

What is PAI? See: PAI Project Overview

What is a Pack? See: Pack System

This pack installs the CORE skill - the brain of PAI. CORE is:

  • Auto-loading - Activates at every session start
  • Authoritative - The single source of truth for how PAI operates
  • Foundational - Other skills depend on CORE for configuration and routing

Without CORE, your PAI system has no:

  • Response format standards
  • Workflow routing
  • System architecture documentation
  • Security protocols
  • Memory system integration

Core principle: CORE is the operating system of PAI - everything else builds on it.

Please follow the installation instructions in INSTALL.md to integrate this pack into your infrastructure.

What's Included

Component Location Purpose
SKILL.md skills/CORE/SKILL.md Main skill definition with routing and configuration
SYSTEM docs skills/CORE/SYSTEM/ 19 architecture and system documentation files
USER templates skills/CORE/USER/ Empty user customization structure with READMEs
WORK templates skills/CORE/WORK/ Sensitive work directory placeholder
Workflows skills/CORE/Workflows/ 4 core workflows (Delegation, SessionContinuity, etc.)
Tools skills/CORE/Tools/ 4 CLI tools (Inference, SessionProgress, etc.)

Summary:

  • Files created: 34
  • SYSTEM docs: 19 architecture files
  • Workflows: 4
  • Tools: 4
  • Dependencies: None (foundation pack)

The Concept and/or Problem

AI agents are powerful but lack structure. Without a foundational system:

Without Response Standards:

  • Every response has different format
  • No consistent way to communicate
  • Voice integration impossible
  • Users can't predict what they'll get

Without Architecture Documentation:

  • No single source of truth
  • Configuration scattered everywhere
  • New features break old ones
  • Debugging is guesswork

Without Workflow Routing:

  • User intent to action is fuzzy
  • Same request handled differently each time
  • No way to extend functionality predictably
  • Integration points undefined

Without SYSTEM/USER Separation:

  • Updates overwrite customizations
  • Personal data leaks to public repos
  • No safe way to personalize

The Fundamental Problem:

AI systems need scaffolding - not just capabilities, but the organizational structure that makes capabilities reliable, extensible, and maintainable. CORE provides that scaffolding.

The Solution

CORE solves this through layered architecture with explicit contracts:

Layer 1: Response Format

Every AI response follows a predictable structure:

  • SUMMARY for quick understanding
  • ANALYSIS for findings
  • ACTIONS for what was done
  • RESULTS for outcomes
  • Voice output for TTS integration

Layer 2: SYSTEM/USER Two-Tier

Configuration that never conflicts:

  • SYSTEM: Base defaults, updated with PAI
  • USER: Your customizations, never touched

Layer 3: Workflow Routing

Intent maps to action predictably:

  • Triggers in SKILL.md route to workflows
  • Workflows document exact procedures
  • Tools provide CLI interfaces

Layer 4: Documentation as Code

Architecture docs that stay current:

  • PAISYSTEMARCHITECTURE.md: Founding principles
  • SKILLSYSTEM.md: How skills work
  • MEMORYSYSTEM.md: How history works
  • And 16 more specialized docs

What Makes This Different

This sounds similar to "system prompts" which also configure AI behavior. What makes this approach different?

CORE is not a prompt - it's an operating system. While system prompts are static text that gets prepended to conversations, CORE is a dynamic skill with routing tables, documentation hierarchies, and tool integrations. It loads selectively based on context, routes intent to specific workflows, and maintains separation between system defaults and user customizations.

  • System prompts are static; CORE routes dynamically
  • Prompts overwrite on update; USER tier is protected
  • Instructions are flat; CORE has explicit layer hierarchy
  • Configuration is scattered; CORE centralizes everything

Configuration

Environment variables (add to shell profile or .env):

# Required
export PAI_DIR="$HOME/.claude"

# Optional - for voice integration
export DA="YourAIName"
export TIME_ZONE="America/Los_Angeles"

settings.json configuration:

{
  "daidentity": {
    "name": "YourAIName",
    "fullName": "Your AI Full Name",
    "voiceId": "your-elevenlabs-voice-id"
  },
  "principal": {
    "name": "YourName",
    "timezone": "America/Los_Angeles"
  }
}

Customization

Recommended Customization

Populate your USER directory:

After installing CORE, personalize your PAI by creating files in USER/:

  1. ABOUTME.md - Tell your AI about yourself
  2. BASICINFO.md - Name, timezone, location
  3. DAIDENTITY.md - Customize your AI's personality
  4. RESPONSEFORMAT.md - Override default response format

Why: USER files make PAI yours. Without them, you get generic defaults.

Process:

  1. Read the README.md files in each USER subdirectory
  2. Create the recommended files with your personal content
  3. Your AI will use this context in every session

Expected Outcome: An AI that knows you, speaks in your preferred style, and maintains your preferences across sessions.

Optional Customization

Customization Location Impact
Security patterns USER/PAISECURITYSYSTEM/patterns.yaml Custom sensitive data detection
Skill preferences USER/SKILLCUSTOMIZATIONS/{Skill}/ Per-skill behavior overrides
Banner config USER/BANNER/config.yaml Session start display
Terminal settings USER/TERMINAL/preferences.yaml Terminal appearance

Credits

  • Original concept: Daniel Miessler - developed as the foundation of PAI (Personal AI Infrastructure)
  • Inspired by: Unix philosophy (modular tooling), Anthropic's agent harness patterns, and engineering best practices

Changelog

2.3.0 - 2026-01-14

  • Initial pack release for PAI v2.3
  • Includes complete SYSTEM documentation (19 files)
  • USER directory templates with README guides
  • WORK directory template for sensitive content
  • 4 core workflows (Delegation, SessionContinuity, ImageProcessing, Transcription)
  • 4 CLI tools (Inference, SessionProgress, FeatureRegistry, SkillSearch)
  • CRITICAL: USER directory is empty template - populate with your personal content