Method Kit

AI-first development methodology — progressive autonomy, spike validation, governed writes, and knowledge-as-code.

What is this?

A portable development methodology packaged as installable scaffolding. It gives any AI-assisted project:

• Structured decision records (ADRs) before code
• Spike-based validation with objective green criteria
• Progressive autonomy ladder (A0→A3) for AI agents
• Knowledge-as-code governance
• Classification and capability-based access control
• Demo-as-contract validation
• Mandatory closing blocks for state awareness

Quick Start

npx @method-kit/create my-project

This installs into your current directory:

.method-kit/          — methodology assets (VERSION, scripts)
.agents/skills/       — 5 governance skills for AI agents
.memory/              — living memory (decisions, vision, glossary)
docs/                 — roadmap, spike criteria
AGENTS.md             — agent contract (§1-§8)

What it is

✅ A methodology for developing with AI assistance ✅ Governance framework for AI agent autonomy ✅ Templates for decisions, spikes, roadmaps ✅ Skills that teach AI agents the discipline ✅ Stack-agnostic, model-agnostic, language-agnostic

What it is NOT

❌ A framework or library ❌ A product template ❌ Tied to any specific AI model or tool ❌ A replacement for your build system ❌ An opinionated tech stack

The Seven Pillars

1. Horizon Progression — sequential, evidence-gated
2. Spike Validation — objective green criterion before "done"
3. Decision Before Code — ADR registered before implementation
4. Governed Writes — proportional approval pipeline
5. Classification × Capability — structural access control
6. Progressive Autonomy — A0→A3 with audit trail
7. Demo as Contract — one command proves everything

Full methodology: METHODOLOGY.md

Compatibility

Works with any AI coding tool that reads project files:

• Claude Code / Opus 4.8 (reads AGENTS.md, .mcp.json)
• Cursor (reads .cursorrules)
• GitHub Copilot (reads .github/copilot-instructions.md)
• Codex CLI
• Windsurf
• Any MCP-compatible agent
• Any model via system prompt injection

After Installation

1. Edit .memory/product-vision.md — define your real problem
2. Edit AGENTS.md §4 — define your surfaces
3. Edit docs/roadmap.md — define 3-5 spikes with green criteria
4. Write founding ADRs (D-001..D-003) in .memory/decisions.md
5. Run your first AI-assisted turn — verify §7 closing block renders

Contributing

This methodology evolves through real usage. If you use it and find improvements:

1. Fork the repo
2. Apply the change
3. Document what real execution revealed the improvement
4. Open a PR with before/after evidence

Origin

Extracted from real execution of AKOS (AI Knowledge Operating System) — v0.1 through v0.6, across 19 ADRs, 11 validated spikes, and 35+ MCP tools. Every pillar and operational rule was stress-tested in production development, not theorized.

License

MIT