Jordach Makaya
Back to Shokunin Architecture
CONCEPT PAPER · VERSION 1.1

Shokunin: A Governed Operating Method for AI Work

The complete methodological foundation: bounded authority, durable state, lean context, controlled process, independent challenge, and proof over promises

By Jordach Makaya · Hard Machine LabsVersion 1.1 · July 2026
Inspect Implementation Architecture

Abstract

As AI models gain capability, engineering teams face a paradox: raw generation velocity increases while confidence in multi-step execution drops. Long agent sessions degrade through context rot, unverified assertions, and self-certified completions.

Shokunin is an operating method designed to resolve this paradox. It treats agent autonomy not as unconstrained model generation, but as bounded execution inside explicit protocols.

The Six Foundational Principles

  1. Bounded Authority (PR-1): Freedom of execution inside strict boundaries of permission and role.
  2. Script-Captured Evidence (PR-2): VERIFIED, never DECLARED. Gates read machine-written proof files, not LLM claims.
  3. Role Separation (PR-3): The agent that writes code never reviews or tests its own diff.
  4. Durable State & Lean Context (PR-4): Session memory lives in versioned disk artifacts rather than sprawling prompt context.
  5. Governed Sequential Execution (PR-5): Explicit phase boundaries with deterministic release gates.
  6. Swarm HOLD Pattern (PR-6): Specialist subagents remain idle on disk between turns, eliminating context reloads during iterative repair cycles.

The 9-Zone Topology (Z1–Z9)

Shokunin structures project lifecycles across nine explicit governance zones: Z1 Governance & Intent, Z2 Requirements & Scope, Z3 Architecture & Contracts, Z4 Task Decomposition, Z5 Implementation, Z6 Verification, Z7 Review, Z8 Security, and Z9 Release Gate.