Warning
This wizard creates a SKELETON directory structure only. After running the wizard, you must manually install each pack in order (pai-hook-system → pai-core-install → etc.) to have a working system.
Want a complete working system immediately? Use the Full Release Install instead — it includes all packs pre-configured.
FOR AI AGENTS: You are about to install the PAI bundle skeleton. Run the wizard, then install each pack in order. The wizard alone does NOT create a working system.
Origin: This bundle is extracted from Daniel Miessler's production Kai system - a battle-tested Personal AI Infrastructure that has been refined through daily use since 2024.
MANDATORY: Read this section before installing ANY pack in this bundle.
AI agents systematically create "simplified" versions of pack implementations. Instead of the full 268-line specification, they create 50-line "equivalents." This breaks the system and wastes the user's time.
- EVERY PACK = FULL IMPLEMENTATION - No shortcuts, no simplification
- EVERY FILE SPECIFIED = EVERY FILE CREATED - Count them
- EVERY LINE OF CODE = EVERY LINE OF CODE - Don't summarize
- IF YOU CAN'T DO IT FULLY, STOP AND SAY SO - Don't create partial implementations
For EACH pack you install, confirm:
## Pack: [pack-name] Installation Verification
- [ ] All directories created as specified
- [ ] All files created (counted: X of Y)
- [ ] Code files contain COMPLETE implementations
- [ ] YAML/JSON files contain ALL entries
- [ ] Verification commands passed
- [ ] No "simplified" or "equivalent" versions createdAfter ALL packs are installed, confirm:
## PAI Bundle Installation Complete
### Pack Status
- [ ] pai-hook-system - FULLY installed, verified
- [ ] pai-history-system - FULLY installed, verified
- [ ] pai-core-install - FULLY installed, verified
- [ ] pai-voice-system - FULLY installed, verified (if applicable)
### Integration Status
- [ ] ~/.claude directory structure correct
- [ ] settings.json contains all hook registrations
- [ ] Startup hook fires on new session
- [ ] Claude Code restarted to activate hooks
### Quality Confirmation
- [ ] NO simplified implementations
- [ ] NO missing files
- [ ] NO summarized configurations
- [ ] User informed of required customizationsPresent this completed checklist to the user. If ANY item fails, you are NOT done.
# Fresh install (with backup)
bun run install.ts
# Update existing installation (preserves config)
bun run install.ts --updateThe wizard will:
- Detect existing AI systems - Scans for Claude Code, Cursor, Windsurf, Cline, Aider, Continue
- Show you what it found and explain what will be backed up
- Ask for your confirmation before making any changes
- Create a safety backup of your existing
~/.claudedirectory to~/.claude-BACKUP - Ask a few questions (your name, AI name, timezone, voice preference)
- Install directly to
~/.claude- the standard Claude Code location - Configure environment variables automatically
Safety First: The installer will not modify anything until you explicitly confirm. Your original files are always preserved in the backup.
After pulling new changes from the repo, use --update to apply them without losing your configuration:
git pull
bun run install.ts --updateUpdate mode will:
- Skip backup - Your files stay in place
- Read existing config - Uses your .env values as defaults
- Preserve customizations - Hooks, history, and personal settings untouched
- Update infrastructure - Only refreshes core skill files
Important
The features below are only available after you install the required packs. The wizard alone only creates the directory structure and configuration files.
When fully installed (wizard + all required packs), the PAI bundle gives you:
- A custom History system - Automatically captures sessions, decisions, learnings, and research
- A custom set of Hooks - Event-driven automation that triggers on session start, tool use, and task completion
- A custom Skill system - Modular capabilities that route based on intent
- A custom Voice notification system - Spoken updates via ElevenLabs when tasks complete (optional)
- A custom Agent swarm creation system - Spawn parallel agents with personality templates
- A custom Security control system - Protection against prompt injection and dangerous operations
Caution
The wizard does NOT install these packs for you. You must install them manually using your AI assistant after the wizard completes. Without these packs, you have an empty skeleton — not a working PAI system.
After running the wizard, install these packs IN ORDER:
| # | Pack | Purpose | Dependencies |
|---|---|---|---|
| 1 | pai-hook-system | Event-driven automation | None |
| 2 | pai-core-install | Skills + Identity + MEMORY system | Hooks |
| # | Pack | Purpose | Dependencies |
|---|---|---|---|
| 3 | pai-statusline | 4-mode responsive status line | Core |
| 4 | pai-voice-system | Voice notifications | Hooks, Core |
| 5 | pai-observability-server | Multi-agent monitoring | Hooks |
| Pack | Purpose |
|---|---|
| pai-agents-skill | Dynamic agent composition |
| pai-algorithm-skill | The Algorithm - verifiable iteration |
| pai-annualreports-skill | Security report analysis |
| pai-art-skill | Visual content generation |
| pai-brightdata-skill | Progressive web scraping |
| pai-browser-skill | Browser automation |
| pai-council-skill | Multi-agent debate |
| pai-createcli-skill | CLI tool generation |
| pai-createskill-skill | Skill creation |
| pai-firstprinciples-skill | First principles analysis |
| pai-osint-skill | Open source intelligence |
| pai-privateinvestigator-skill | Ethical people-finding |
| pai-prompting-skill | Meta-prompting system |
| pai-recon-skill | Security reconnaissance |
| pai-redteam-skill | Adversarial analysis |
| pai-research-skill | Multi-source research |
| pai-system-skill | System maintenance |
| pai-telos-skill | Life OS and goals |
Alternative: Use a Full Release which includes all 23 packs pre-configured in a complete
.claude/directory.
Give each pack directory to your DA and ask it to install:
"Install the pai-hook-system pack from PAI/Packs/pai-hook-system/"
The AI will:
- Read the pack's
README.mdfor context - Follow
INSTALL.mdstep by step - Copy files from
src/to your system - Complete
VERIFY.mdchecklist to confirm success
Note: Each pack is now a directory (v2.0 format) containing README.md, INSTALL.md, VERIFY.md, and a src/ folder with actual code files.
- Hooks are the foundation - they enable all event-driven automation
- History uses hooks to capture events and context
- Core Install provides skill routing and identity framework
- Voice uses hooks for completion events (requires ElevenLabs API key)
- Bun:
curl -fsSL https://bun.sh/install | bash - Claude Code or compatible AI coding assistant
After installing all packs:
# Check directory structure
ls -la ~/.claude/
# Expected directories:
# hooks/ - Event-driven automation
# history/ - Sessions, Learnings, Research, Decisions
# skills/ - CORE and other skills
# tools/ - CLI utilities
# voice/ - Voice server files (if installed)
# Check hooks are registered
cat ~/.claude/settings.json | grep -A 5 "hooks"
# Restart Claude Code to activate all hooksIf something goes wrong:
# Remove the new installation
rm -rf ~/.claude
# Restore from backup
mv ~/.claude-BACKUP ~/.claudePacks are complete subsystems organized around a single capability. Each pack is a directory containing:
README.md- Overview, architecture, what it solvesINSTALL.md- Step-by-step installation instructionsVERIFY.md- Mandatory verification checklistsrc/- Actual source code files (TypeScript, YAML, etc.)
For example, pai-hook-system provides an entire event-driven automation framework with all code files ready to copy.
Bundles are curated combinations of packs designed to work together. The Official PAI Bundle is 10 packs that form a complete AI infrastructure (4 required + 6 optional).
The PAI system embeds these principles from Personal AI Infrastructure:
- Clear Thinking + Prompting is King - Good prompts come from clear thinking
- Scaffolding > Model - Architecture matters more than which model
- As Deterministic as Possible - Templates and consistent patterns
- Code Before Prompts - Use AI only for what actually needs intelligence
- Spec / Test / Evals First - Write specifications and tests before building
- UNIX Philosophy - Do one thing well, make tools composable
- ENG / SRE Principles - Treat AI infrastructure like production software
- CLI as Interface - Command-line is faster and more reliable
- Goal -> Code -> CLI -> Prompts -> Agents - The decision hierarchy
- Meta / Self Update System - Encode learnings so you never forget
- Custom Skill Management - Modular capabilities that route intelligently
- Custom History System - Everything worth knowing gets captured
- Custom Agent Personalities - Different work needs different approaches
- Science as Cognitive Loop - Hypothesis -> Experiment -> Measure -> Iterate
- The Algorithm: 7-phase problem-solving with ISC tracking for verifiable iteration
- 29 Skills: Up from 23, expanded domain coverage
- 15 Hooks: Refined hook architecture
- 331 Workflows: Comprehensive automation library
- Improved Wizard: Renamed to PAIInstallWizard.ts, better interactive setup
- Voice Notifications: ElevenLabs integration for spoken alerts
- Observability Dashboard: Real-time system monitoring
- Full Releases Return: Added Releases/v2.3/ with complete .claude directory
- 23 Packs Total: 5 infrastructure + 18 skill packs
- Continuous Learning: Sentiment capture, rating system, learning extraction
- New Packs: Added pai-statusline, pai-annualreports-skill, pai-brightdata-skill, pai-council-skill, pai-createcli-skill, pai-createskill-skill, pai-firstprinciples-skill, pai-osint-skill, pai-privateinvestigator-skill, pai-recon-skill, pai-redteam-skill, pai-research-skill, pai-system-skill, pai-telos-skill
- History Retired: MEMORY system in pai-core-install replaces standalone history pack
- Rebranded to PAI: All packs renamed from pai-* to pai-* for consistency with project name
- New Pack: Added pai-upgrades-skill for tracking Anthropic ecosystem updates
- New Pack: Added pai-browser-skill for web automation and verification
- Bundle renamed: "Kai Bundle" -> "Official PAI Bundle"
- Origin preserved: Bundle documentation notes extraction from Daniel Miessler's Kai system
- Directory-based Packs: All packs migrated from single markdown files to directory structure
- New Pack Format: Each pack now contains README.md, INSTALL.md, VERIFY.md, and src/ directory
- Actual Code Files: Source code now in real .ts, .yaml, .hbs files (not embedded in markdown)
- Four New Packs: Added pai-prompting-skill, pai-agents-skill, pai-art-skill, pai-observability-server
- Improved Installation: AI agents copy actual files instead of extracting from markdown blocks
- AI System Detection: Scans for Claude Code, Cursor, Windsurf, Cline, Aider, Continue
- Clear Communication: Shows exactly what was detected and what will be backed up
- Explicit Confirmation: Asks permission before making any changes to your system
- Safety-first approach: No modifications until user confirms
- Now installs directly to
~/.claudeinstead of configurable$PAI_DIR - Automatic backup to
~/.claude-BACKUPbefore installation - Environment variables set automatically (no manual shell sourcing needed)
- Simplified: Removed personality calibration questions - just name, timezone, voice
- Simplified: Removed technical preference questions - use sensible defaults
- Simplified: Removed "open another terminal" instructions
- Initial release with full wizard
Author: Daniel Miessler Origin: Extracted from production Kai system (2024-2026)