microhard

MicroHard Meta-Framework

CAI (Client Acquisition Intelligence) + MGI (Meta-General Intelligence)
A step-driven workflow engine with cross-session persistence and intelligent agent orchestration.

🚀 What is MicroHard?

MicroHard is a meta-framework for AI-driven workflows that combines:

• CAI (Client Acquisition Intelligence) — B2B lead generation, content creation, and growth optimization
• MGI (Meta-General Intelligence) — Strategic thinking, mental models, adversarial design, and knowledge synthesis

Unlike traditional single-session AI tools, MicroHard provides cross-session persistence, step-based workflows, and intelligent knowledge libraries that work across days and weeks.

Core Architecture

• ✅ Step-driven state machine — Focused execution, one step at a time
• ✅ Cross-session persistence — Resume workflows after days or weeks
• ✅ Intelligence Layer — 99-model mental models library with quick-select index
• ✅ Workflow types — Linear, selection, iterative, and exploratory patterns
• ✅ Data injection — Skills carry domain knowledge that's auto-injected at runtime
• ✅ 24 specialized skills — 6 MGI + 18 CAI domain experts

🛠 Features

Workflow Engine

• 4 Workflow Types: Linear, selection, iterative, and exploratory
• Loop Support: --loop flag returns to checkpoint for iterative refinement
• Step Navigation: --step <id> jumps to any step in exploratory workflows
• State Persistence: Three-layer state system (state.json, context.md, project.state.json)

Intelligence Layer (MGI)

• 99 Mental Models: Curated knowledge library with quick-selection index
• 6 MGI Skills: Theory practitioner, adversarial designer, first principles, cross-domain synthesis, system auditor, knowledge curator
• Data Injection: Reference data automatically loaded per step
• Selection Workflows: Choose the right mental model, then apply it

Client Acquisition (CAI)

• 18 CAI Skills: Hook engineering, content architecture, ad creative, growth optimization
• 4-Phase Pipeline: Attract → Convert → Scale → Optimize
• Performance Tracking: Built-in compliance, feedback loops, and CEO reporting

📦 Installation

Package: microhard on npm
Prerequisites: Node.js v20+

Quick Install (Recommended)

npx microhard install

Or install globally if you'll use it across many projects:

npm install -g microhard
microhard install

What Gets Installed?

The installer will guide you to:

• Select AI IDEs — Claude Code, Gemini CLI, GitHub Copilot
• Select Modules — MGI skills, CAI skills, or both
• Install Skills — Copies to .claude/skills/, .gemini/skills/, .github/skills/

📖 Full installation guide: INSTALLATION.md

For Development

# Clone repository
git clone <repository>
cd MicroHard

# Install dependencies
npm install

# Run quality checks
npm run quality

# Validate skills
npm run validate:strict

# Install skills locally
node tools/installer/microhard-cli.js install

Quality Control Setup

MicroHard includes built-in quality control for AI-generated code:

• ESLint — Prevents common bugs in JavaScript
• Prettier — Ensures consistent formatting
• Markdownlint — Validates markdown (your source code!)

npm run quality        # Check all code quality
npm run quality:fix    # Auto-fix issues

Pre-commit hooks automatically run before each commit. See docs/QUALITY-CONTROL.md for details.

🚦 Quick Start

After installation, skills are available as slash commands in your AI IDE:

/mgi-theory-practitioner
/mgi-adversarial-designer
/cai-hook-engineer
/cai-content-architect

Example Usage

In Claude Code / Gemini CLI / Copilot:

You: /mgi-theory-practitioner

AI: I'm the MGI Theory Practitioner. I'll help you select and apply
    the right mental model to your problem.

    Please describe the problem you're trying to solve.

You: I need to improve my SaaS pricing strategy.

AI: [Loads mental models library and suggests relevant frameworks...]

Using the CLI (Optional)

# View progress across all skills
microhard status

# Run a skill (show current step)
microhard run cai-hook-engineer

# Mark step done and advance
microhard run cai-hook-engineer --next

# For iterative workflows, loop back to refine
microhard run mgi-adversarial-designer --loop

# For exploratory workflows, jump to any step
microhard run mgi-cross-domain-synthesizer --step step-03-synthesis

# Rebuild mental models index
microhard library index

📖 See: docs/SLASH-COMMANDS-REFERENCE.md for all 24 skills

📚 Documentation

Architecture & Design:

• MicroHard Architecture Design — Complete architectural specification
• Implementation Status — Phase completion tracking and validation

Skills Reference:

• CAI Skills — 18 content acquisition and growth optimization skills
• MGI Skills — 6 strategic thinking and knowledge synthesis skills
• Mental Models Library — 99 indexed mental models

🧱 Project Structure

src/
├── cai-skills/           # Client Acquisition Intelligence (18 skills)
│   ├── 1-attract/        # Hook engineering, content architecture
│   ├── 2-convert/        # Ad creative, landing pages
│   ├── 3-scale/          # Growth optimization, A/B testing
│   └── 4-optimize/       # Feedback, compliance, CEO reporting
│
├── mgi-skills/           # Meta-General Intelligence (6 skills)
│   ├── 1-distillation/   # Theory practitioner, knowledge curator
│   ├── 2-stress-test/    # Adversarial designer, first principles
│   ├── 3-synthesis/      # Cross-domain synthesis, system auditor
│   └── _library/         # Mental models library (99 models + index)
│
├── cai-workflows/        # Multi-skill orchestration workflows
└── templates/            # Client project boilerplates

tools/
├── engine/               # Workflow execution engine
│   ├── workflow-runner.js    # Step orchestration, loop/navigation
│   ├── state-manager.js      # Cross-session persistence
│   └── registry-reader.js    # Skill resolution
│
├── installer/            # Installation and CLI
│   └── microhard-cli.js      # Main CLI interface
│
└── library/              # Library management
    └── build-model-index.js  # Mental models index generator

🎯 Workflow Types Explained

Type	Description	Example Skills	Use Case
Linear	Sequential steps, executed in order	All CAI skills, mgi-knowledge-curator	Standard workflows with clear prerequisites
Selection	Choose from library, then apply	mgi-theory-practitioner	Select the right tool, then use it
Iterative	Loop until quality threshold met	mgi-adversarial-designer, mgi-first-principles-deconstructor	Refinement cycles with checkpoints
Exploratory	Non-linear, jump between steps	mgi-cross-domain-synthesizer	Discovery-driven workflows

🧠 MGI Skills Overview

Skill	Type	Purpose
Theory Practitioner	Selection	Apply mental models from library to real problems
Adversarial Designer	Iterative	Attack plans, audit attacks, harden against them
First Principles Deconstructor	Iterative	Strip to axioms, rebuild from ground truth
Cross-Domain Synthesizer	Exploratory	Transfer insights across unrelated domains
System Auditor	Iterative	Audit systems with domain-specific lenses
Knowledge Curator	Linear	Ingest raw material into mental models library

📊 CAI Skills Overview

Phase 1 — Attract:

• Hook Engineer, Content Architect, Topic Matrix Generator

Phase 2 — Convert:

• Ad Creative Engineer, Landing Page Architect

Phase 3 — Scale:

• Growth Controller, A/B Test Designer

Phase 4 — Optimize:

• Feedback Loop Tracker, Compliance Auditor, CEO Architect

🔬 Implementation Status

✅ Phase 1 Complete: Library Index Generator (99 models indexed)
✅ Phase 2 Complete: Data Injection Layer (11 data folders)
✅ Phase 3 Complete: Workflow Type System (4 types implemented)
✅ Phase 4 Complete: MGI Skill Redesigns (all 6 skills updated)

See IMPLEMENTATION-STATUS.md for detailed progress tracking.

🤝 Contributing

MicroHard is built for extensibility. To contribute:

• Read the Architecture Design
• Review Implementation Status
• Follow the skill anatomy patterns in existing skills
• Run node tools/validate-skills.js before submitting

📄 License

MIT License — Built with 🧠 for systematic architects and strategic thinkers.

MicroHard Meta-Framework
Where intelligence meets execution.