The Architect's Loop

Five moves. One dependency order.

Find the first layer your workflow has not locked in. Build one inspectable artifact there before adding another agent, automation, or metric.

Assessment → artifact

The method ends in a buildable handoff.

01

Assess

Answer five dependency questions locally in the browser.

02

Export

Download or copy the generated Markdown architecture brief.

03

Build

Use only the starter artifact for the first move that is not locked in.

04

Review

Keep, revise, or stop the system using its acceptance evidence.

Ordered dependencies

Build the first missing layer—not all five.

01

See

Can the system find the decisions and constraints it needs?

Build
A scoped memory index
Accept when
An agent can find the latest decision and cite its source.
Guardrail
Exclude secrets and private material the workflow does not need.
02

Design

Is the repeating job written before execution begins?

Build
A one-job system specification
Accept when
Another person can explain the input, output, owner, boundary, and stop condition.
Guardrail
Do not automate an ambiguous job.
03

Build

Can one bounded worker complete one part reliably?

Build
A single-purpose agent or skill
Accept when
The worker passes one representative fixture and fails safely on an invalid one.
Guardrail
Limit permissions and tools to the named job.
04

Automate

Are trigger, review, failure, and receipt connected?

Build
A supervised workflow loop
Accept when
One run produces an output, review state, failure path, and durable receipt.
Guardrail
Keep consequential actions human-approved.
05

Compound

Can two runs be compared using the same outcome evidence?

Build
A feedback and review protocol
Accept when
The next run records what changed and why against the same evidence-backed score.
Guardrail
Do not let proxy metrics silently replace the real outcome.

Public boundary

A workflow method—not a diagnosis or promise.

The assessment evaluates workflow architecture from the answers you choose. It does not evaluate personality, mental or physical health, financial potential, or future outcomes.

Private inputs stay under your control.

Optional assessment context remains in browser state. The generated brief is visible only to you unless you choose to download, copy, or share it.

Find the gap, then open the matching build path.

The assessment produces the scope, acceptance test, guardrail, and seven-day order. Implementation options remain separate from the free method and are not a checkout promise.