What is MicroHard?

MicroHard is an AI CLI meta-framework that turns your AI assistant into a structured organizational brain. You install a set of skills — slash commands — each representing a specialized domain expert. The AI becomes that expert for the duration of the session.

There are no servers, no APIs, no cloud services. The AI (Claude, Gemini, Copilot) is the execution engine.

/mh-cto-sdlc-prd-create       → AI becomes your Product Definition Architect
/mh-ceo-pricing-architecture  → AI becomes your Pricing Architect
/mh-coo-quality-assurance     → AI becomes your Quality Auditor
/mh-mgi-theory-practitioner   → AI applies mental models to your problem

Skills have cross-session memory — decisions persist across conversations, context survives overnight, multi-step engagements resume where they left off.

Departments

MicroHard ships as four C-suite departments:

| Department | Focus | Skills | | ---------- | --------------------------------------- | ------------------------- | | CEO | Strategy, positioning, growth, pricing | 13 skills | | CTO | Full SDLC — ideation through validation | 60 skills across 9 phases | | CMO | Lead generation, pipeline | 2 skills (expanding) | | COO | Delivery, quality, team, knowledge ops | 16 skills |

Plus MGI (Meta-General Intelligence) — 6 cross-domain thinking tools built on a 99-model mental models library.

Total: 109 skills.

Installation

Prerequisites: Node.js v20+

Run the installer from a parent directory — it will create a new folder for your project name and install everything inside it.

npx microhard install
# then:
# cd <your-project-name>

Or globally:

npm install -g microhard
microhard install
# then:
# cd <your-project-name>

The installer asks:

1. Project name and your name — used to name the project folder and personalize the context file
2. Departments — select CEO, CTO, CMO, COO (any combination)
3. MGI module — the mental models library (recommended)
4. AI IDEs — Claude Code, Gemini CLI, GitHub Copilot (any combination)

Skills are installed as slash commands:

• Claude Code → .claude/skills/{skill-name}/
• Gemini CLI → .gemini/skills/{skill-name}/
• GitHub Copilot → .github/skills/{skill-name}/

After install, cd into your new project folder and type /mh- in your AI IDE to see all available skills.

Architecture

Three Namespaces

src/
├── core/                           # Domain-agnostic skills (mh-{skill})
│   ├── mh-brainstorming/
│   ├── mh-distillator/
│   └── ...
│
├── departments/                    # C-suite domain skills
│   ├── ceo/                        # mh-ceo-{sub}-{skill}
│   │   ├── business-strategy/
│   │   ├── client-strategy/
│   │   ├── growth-strategy/
│   │   └── service-design/
│   ├── cto/
│   │   └── sdlc/                   # 9 phases, mh-cto-sdlc-{skill}
│   │       ├── phase-0-ideation/
│   │       ├── phase-1-analysis/
│   │       └── ... (through phase-8)
│   ├── cmo/
│   └── coo/
│       ├── client-ops/
│       ├── delivery-ops/
│       ├── knowledge-ops/
│       └── team-ops/
│
└── modules/
    └── mgi/                        # mh-mgi-{skill}
        └── _library/               # 99 mental models

Project Infrastructure (_mh/)

After install, your project gets:

_mh/
├── config.toml           # Project config (name, user, installed depts)
├── _config/
│   └── skill-manifest.csv  # Searchable index of all installed skills
├── state/
│   └── project.state.json  # Completed skill tracking
└── custom/               # Your overrides (survives updates)

_mh-output/               # All skill outputs land here
├── ceo/
├── cto/
│   └── sdlc/
│       ├── prd/
│       ├── sprint-status.yaml
│       └── ...
├── cmo/
└── coo/

mh-project-context.md     # Fill this in — AI reads it at every session start

Cross-Session Memory

The AI reads mh-project-context.md at the start of each skill invocation. This file is your persistent briefing — project name, stack, active sprint, constraints. Fill it in once; every skill invocation inherits that context without re-explaining.

CTO / SDLC — 9 Phases

The CTO department covers the complete software delivery lifecycle:

| Phase | Skills | Purpose | | ---------------------- | ------------------------- | --------------------------------------------------------- | | 0 — Ideation | mh-cto-sdlc-phase0-gate | Validate the idea before building | | 1 — Analysis | 7 skills | User research, risk, domain model, stakeholder map | | 2 — Product Definition | 6 skills | PRD, epic map, story spec, acceptance criteria | | 3 — UX Design | 9 skills | Wireframes, flows, design system, handoff covenant | | 4 — Architecture | 9 skills | System decomp, tech selection, data arch, security | | 5 — Testing Strategy | 8 skills | Test prioritization, contract/chaos/performance tests | | 6 — Implementation | 6 skills | Sprint planning, TDD sessions, code review, story kickoff | | 7 — Launch | 5 skills | Production readiness, launch plan, go-live, hypercare | | 8 — Validation | 7 skills | Outcome measurement, tech retro, roadmap refresh |

Skills enforce dependency order — gate skills block progress until prerequisites are satisfied.

Quick Start

After npx microhard install, open your AI IDE and type any skill:

/mh-ceo-pricing-architecture

/mh-cto-sdlc-prd-create

/mh-coo-quality-assurance

/mh-mgi-theory-practitioner

The AI will adopt the agent identity, load its domain knowledge, and run through the structured workflow.

Optional CLI

microhard status                       # Project progress across all skills
microhard run mh-cto-sdlc-prd-create  # Show current step
microhard run mh-cto-sdlc-prd-create --next  # Advance to next step
microhard run mh-cto-sdlc-prd-create --reset # Restart from step 1

Skill Anatomy

Every skill is a directory with:

mh-ceo-pricing-architecture/
├── SKILL.md                  # Identity, description, frontmatter
├── customize.toml            # Configurable parameters
├── mh-skill-manifest.yaml   # Dependencies, metadata, phase
└── steps/
    ├── step-01-context.md
    ├── step-02-analysis.md
    └── ...

The AI reads SKILL.md when you invoke the slash command, takes on the agent identity described there, and walks through the steps.

Validation

npm run validate          # Check all 109 skills (warning mode)
npm run validate:strict   # Treat HIGH+ findings as errors
npm run quality           # ESLint + Prettier + Markdownlint
npm run quality:fix       # Auto-fix all quality issues

Documentation

• Complete Guide — Full architecture, all 109 skills, output structure, config system, dependency system, developer guide
• Installation — Step-by-step install and setup
• Quality Control — ESLint, Prettier, Markdownlint setup

License

MIT — Built for solo founders and agency operators who need a systematic AI workflow, not a chat session.