REFERENCE / EXECUTION CONTROL

Execution control, measured by one test.

As used here, execution control concerns whether a requested digital action becomes operational, institutional, financial, legal or public consequence. Many controls address part of that question. Execution Control Infrastructure is the category defined by the full test: the independent control layer that governs the Execution Boundary and determines whether a consequential action is cleared to proceed.

Execution Control Infrastructure · Human Authority Layer · Execution Boundary · Control Before Consequence.

Seven questions decide whether a control is Execution Control Infrastructure.

Criteria for Execution Control Infrastructure · OATHOR LTD
IDQuestionYesNo
ECI-1Is the determination made outside the system that takes the action?
ECI-2Is it beyond the reach of the actor, so that the actor cannot determine, alter, satisfy, bypass or substitute the conditions of its own clearance?
ECI-3Is it free of control by the function or provider that operates or supplies the acting system?
ECI-4Is it made before consequence exists, while the action is still preventable?
ECI-5Does it decide whether this specific action is cleared to become consequence, or only whether the actor is permitted?
ECI-6Is it applied at the Execution Boundary as a matter of course, not only as an occasional review or exception?
ECI-7Is the record created at the moment of determination, rather than reconstructed afterwards?

A control that answers yes to all seven meets the criteria for Execution Control Infrastructure. A control that answers no to any of the first three remains inside the authority of the system it governs.

Criteria first set out in the Foundation Paper, Version 1.0, May 2026. ECI-2 and ECI-3 follow the definition of independence published 24 September 2026.

How each question is applied.

These application notes explain how ECI-1 to ECI-7 are applied. They do not modify, replace or extend the criteria. Where there is any inconsistency, the published criteria control. The notes are read with the definitions of the Execution Boundary, Independent Clearance Determination and Independence, and do not prescribe any product or implementation.

Applying all seven questions

Is the determination made outside the system that takes the action?

The system that takes the action includes every component that requests, prepares or carries out the action, such as models, agents, workflows, orchestration, tools, memory and the runtime in which they execute, and any person who acts as or for that system. A component does not become outside that system by being separately named, separately deployed or described as a separate layer. A control does not become part of that system only because the action passes through it.

Is it beyond the reach of the actor, so that the actor cannot determine, alter, satisfy, bypass or substitute the conditions of its own clearance?

The actor includes the acting system, anyone who requests the action or directs the acting system in preparing or carrying it out, and instructions or content the acting system receives. A function independent of the acting system that sets the conditions is not the actor. The answer is no where the actor, by virtue of being the actor, can determine, alter, satisfy, bypass or substitute the conditions of its own clearance, including where an action can proceed without the control being applied or when no determination is made.

Is it free of control by the function or provider that operates or supplies the acting system?

Control means the practical ability to set, change, suspend or override the conditions or the outcome of the determination. A relationship alone is not control: an organisation or provider may host, integrate, support or supply technology used in the arrangement without controlling the determination. The question applies to the internal function that operates the acting system and to its external provider, where either holds that ability. A provider that neither operates nor supplies the acting system is not within this question. Whether an arrangement meets the criteria depends on the actual allocation of authority and control, not on who supplied the technology.

Is it made before consequence exists, while the action is still preventable?

Consequence includes any operational, institutional, financial, legal or public effect of the action. A preparatory step that has such an effect outside the acting system, and that refusing the action cannot undo, such as data disclosed to a third party or a commitment communicated, is a consequential action in its own right, to which the questions apply, and the assessor records the control relied upon for it. Where a single action is carried out in several steps, the determination precedes the first step that is consequence.

Does it decide whether this specific action is cleared to become consequence, or only whether the actor is permitted?

The determination concerns the specific action, with its actual parameters, at the time it is to proceed. A permission, approval, mandate, policy or session covering a class of actions does not by itself answer this question. The determination must resolve whether the action that becomes consequence is cleared.

Is it applied at the Execution Boundary as a matter of course, not only as an occasional review or exception?

As a matter of course means that every consequential action of each type for which the control is relied upon (G1) reaches the determination, without depending on sampling, on the actor’s choice or on an exception being detected. A label such as low-risk or routine does not by itself take an action outside the questions where the action otherwise becomes consequence, whoever applies the label. Any exclusion by value, type or other condition is recorded with the answer. Where an excluded consequential action falls within the class of actions for which the control is relied upon, the answer is no.

Is the record created at the moment of determination, rather than reconstructed afterwards?

The record must be created contemporaneously with the determination and must identify the determination sufficiently to show that it existed at that time.

Adjacent controls, mapped to the Execution Boundary.

Each term below is defined by what it does and where it sits. Any of these controls qualifies as Execution Control Infrastructure only where it independently determines, at the Execution Boundary and before consequence, whether the specific action is cleared to proceed.

On the system, before software runs

Application execution control

Determines which software, commands or scripts may run on a system.

Answers: Which software may execute, not whether a specific consequential action is cleared to proceed.

Before the Execution Boundary

Identity and authentication

Establish who or what is acting.

Answers: Who is acting.

Before the Execution Boundary

Authorisation and permissions

Establish what an actor is allowed to do.

Answers: What an actor may do, not whether this specific action is cleared to proceed.

Before the Execution Boundary

Approval workflows

Route an action to a designated approver before it proceeds.

Answers: Whether an action was approved. Approval does not by itself constitute clearance.

During execution

Runtime authorization

Decides, at the time of the request, what an actor may do.

Answers: What an actor may do now. Where the actor or its provider controls it, it does not meet the independence criterion.

Within or around the acting system

Guardrails

Shape what a model or agent generates and attempts.

Answers: Whether behaviour stays within defined limits.

Before or during execution

Policy enforcement

Applies an organisation’s policy to requests and actions.

Answers: Whether a request conforms to what is required.

During execution

Runtime governance

Applies policy while a system or agent runs.

Answers: Whether running behaviour conforms to policy.

Within the agent platform

Action gates and agent control planes

Govern an agent’s actions within the platform that runs it.

Answers: What the platform allows its agents to do.

Between agents and the systems they act on

Execution control planes

Enforce policy on agent actions at the point of execution, separately from the agent framework.

Answers: Whether an agent’s action is permitted under policy. It meets the test only where the determination is also free of control by the function or provider that operates or supplies the acting system.

Around the acting system

Fraud and risk controls

Assess the likelihood of loss, error or misuse.

Answers: How likely a source of loss is.

Across the system lifecycle

Human oversight

Supervision of a system’s operation by accountable people.

Answers: What accountable people can see and supervise. Oversight can observe. Authority must still be able to affect execution.

During and after execution

Monitoring and observability

Observe what occurs.

Answers: What is happening.

After consequence

Audit

Records what has occurred.

Answers: What happened.

At the Execution Boundary

Execution Control Infrastructure

Execution Control Infrastructure is the independent control layer that governs the Execution Boundary and determines whether a consequential action is cleared to proceed.

Answers: Whether this specific consequential action is cleared to proceed, independently and before consequence. Only a cleared action proceeds.

Execution Control Infrastructure

Each control is necessary. As defined, none makes the independent determination.

The test decides, not the label.

Any control, under any name, can be measured by the seven questions. A control that answers yes to all seven meets the criteria for Execution Control Infrastructure, whatever it is called. A control that answers no to any of them is an adjacent control, whatever it is called.

New terms do not create new positions at the Execution Boundary. They take a position against the same test.

Permissioned, authenticated, approved or prepared does not necessarily mean cleared to become consequence.

Reference record. Author: OATHOR LTD. Canonical URL: https://oathor.com/execution-control.