Intent
State the outcome, success signals, and deliberate non-goals.
Core method · v0.1
Map the operating logic before accelerating it.
An Agency Map exposes the assumptions that ordinary workflow diagrams leave hidden. It is useful before designing policy, automation, organizational authority, or agent tools.
Actors
Name people, teams, services, agents, and external parties by role.
Authority
Define who may decide, approve, execute, override, or stop—and within what scope.
Boundaries
Describe prohibited outcomes and the mechanism that enforces each limit.
Trust signals
Identify the current, scoped evidence required before action.
Evidence
Preserve durable proof of what happened, why, and under whose authority.
Escalation
Create a visible route for insufficient authority, confidence, or conditions.
Feedback
Assign ownership and cadence to the signals through which the system adapts.
How to use it
Begin with the intended outcome rather than the current process. Describe the minimum system that could produce that outcome safely. Then compare it with reality: misplaced authority, unstated boundaries, stale trust, missing evidence, and informal escalation become visible.
What it is not
An Agency Map is not a permission matrix, a process diagram, or a replacement for domain expertise. It is the common layer that lets those artifacts agree about who can do what, under which conditions, and how the system learns.
Automation should consume an Agency Map—not invent one implicitly.
Machine contract
The experimental JSON Schema is published in the public canon. It supports typed actors, scoped authority, enforceable boundaries, freshness-aware signals, explicit escalation, and owned feedback.