← Back to overview
the framework

Lean architecture for AI.

Treat the context window like a factory workbench. Only what the current operation needs. Everything else is inventory waste.

CONSTITUTION.md hard stops · 30-50 lines always loaded MANIFEST.md routing table · 10-15 lines always visible pulled on demand .context/
the standard

Three files. One contract.

Every repository gets the same structure. The constitution prevents damage. The manifest enables discovery. The supply provides depth. Nothing loads unless earned by the task.

01

Constitution

Always loaded. Contains only what, if missed once, causes irreversible damage. Hard stops. Prohibitions. 30–50 lines. If it exceeds 100, it contains reference material masquerading as rules.

02

Manifest

Always visible. A routing table that maps task patterns to context sources. The agent sees what context exists without holding any of it. 10–15 lines. Never contains the context itself.

03

Supply

Pulled on demand. Individual, focused documents. Each small enough that loading it is cheap, and specific enough that all of it is relevant. Organized by retrieval key, not by chronology.

your-repo/
├── CONSTITUTION.md          ← hard stops, always loaded
│   ┊  "Do not push to main without verifying the build succeeds."
│   ┊  "Do not modify production data without explicit approval."
│   ┊  30-50 lines. Every line load-bearing.
│
├── MANIFEST.md              ← routing table, always visible
│   ┊  - Network changes    → .context/network-standards.md
│   ┊  - Auth modifications  → .context/auth-architecture.md
│   ┊  - Deploy procedures   → .context/deploy-procedures.md
│   ┊  10-15 lines. Task → context source.
│
├── CLAUDE.md                → CONSTITUTION.md (symlink for Claude Code)
├── AGENTS.md                → CONSTITUTION.md (symlink for Copilot)
│
├── .context/                ← supply, pulled on demand
│   ├── auth-architecture.md
│   ├── network-standards.md
│   └── deploy-procedures.md
│
└── src/
four rules

The principles that govern it.

W

The Workbench Rule

Only what the current operation requires should be in the context window. Everything else is inventory waste. Context waste does not simply take up space — it adds noise to signal.

S

The Shadow Board Rule

The agent should see what context is available without holding it. A shadow board in a factory shows the outline of every tool. The manifest is the agent equivalent.

P

The Pull Rule

Context loads when the task demands it, not when the session starts. Push-based context is overproduction. Pull-based context is lean. Pull also makes reasoning visible.

A

The Andon Rule

Stop when uncertain. Surface what is missing. In lean manufacturing, the andon cord stops the production line. The act of stopping is the system working, not failing.

Built on lean manufacturing.

Lean manufacturing solved a structurally identical problem decades ago. A cluttered workstation slows the worker — not because the worker is less skilled, but because the clutter competes for attention and physical space. The solution was not bigger workstations. It was 5S: sort, set in order, shine, standardize, sustain.

The Lean Context Standard is what happens when those principles are applied to the cognitive workbench. The constitution is the sorted workspace — only load-bearing rules survive. The manifest is the shadow board — everything in its place, visible at a glance. The supply is just-in-time delivery — context arrives when the operation demands it. The four rules are the operational discipline that keeps it clean.

We did not copy 5S. We absorbed it. The three tiers and the four rules are what lean looks like when the workbench is a context window and the worker is an agent.

factory floor 5S shadow board just-in-time andon cord context window constitution manifest supply andon rule

This is not an analogy. It is a falsifiable claim: the principles that govern efficient physical workspaces also govern efficient cognitive workspaces. A factory floor and a context window are both finite spaces where work happens, and both degrade in similar ways when overloaded.

Inferable vs. invisible knowledge.

Agents are effective at discovering structure on their own — file layout, code conventions, dependency relationships, API patterns. This is inferable knowledge. Putting it in instruction files is redundant at best and harmful at worst.

What agents cannot discover is invisible knowledge — why a decision was made, what was tried and rejected, what external constraint exists that the code does not encode.

The constitution holds invisible knowledge that prevents damage. The supply holds invisible knowledge that enables depth. Inferable knowledge belongs in neither.

inferable invisible file layout code style dependencies API patterns stays out why this arch what was tried external constraints hard-won context goes in

The strongest configuration is hybrid.

Push-based context wins for narrow, well-defined tasks where the full scope is known upfront and the context fits on the bench. A five-line linting rule set. A fixed coding standard. A short safety checklist.

Pull-based context fails when the manifest is poorly organized, when supply documents are stale, or when the agent lacks the judgment to recognize what it needs.

hybrid model constitution — always pushed manifest — always visible everything below: pulled on demand auth network deploy

The constitution is always pushed — it is small enough and critical enough to justify the cost. The manifest is always visible — it is cheap. Everything else is pulled. This is not a pure pull system. It is a system that pushes only what earns the right to be pushed.

Structure enables trust.

When the system is lean, the agent reasons well. Trust builds. Calibration improves. Both sides get smarter.

The Human Impact →