Decision turns Signals into run, wait, or stop states so the right human, agent, or policy path can act with context.
- Run
- Wait
- Stop
- Escalate
Decision gives operators one surface to review Signals, approve bounded action, ask for more context, or stop work before risk moves downstream.
Decision turns Signals into run, wait, or stop states so the right human, agent, or policy path can act with context.
Atlas shows where the Decision sits, which systems are affected, and what Proof must be recorded after the action.
Decision routes whether the next step can run, must wait, or should stop with a named owner.
A watched source says something changed or needs attention.
The queue routes the update to a human, agent, policy rule, or stop point.
The map keeps the owner, system boundary, and downstream impact visible.
The decision produces evidence, outcome, receipt, audit trail, and rollback note.
In production, Decision is required because it owns review owner, run-wait-stop state, allowed action, escalation and policy reason.
Decision owns the named review owner, allowed action, escalation reason, policy reason, and current run-wait-stop state.
Decision should not start from a vague task. It consumes a Signal with source context, affected system, and reason to review.
Decision produces the minimum proof needed for the next operator to understand what happened and why.
Each page describes one product surface, but production workflows need the four surfaces attached to the same map: Atlas, Signal, Decision, and Proof.
Start with one live workflow, connect the source signal, name the decision owner, and decide which proof record must survive the action.
Decide what a reviewer can approve, reject, route, request, or stop.
Turn hidden judgment into a policy rule or owner approval path.
Make the queue ask for the proof record before the workflow moves on.