Glossary

This page defines the terms used across The Agentic Foresight Registry, so that a term encountered on one page can be checked without leaving the site to guess at it. Some of the terms below describe the product; others describe the method that computes it. Each definition is given once, here, and is not restated with variation elsewhere.

Get AssessmentThe AFR-1 method →

An open reference volume on a glass lectern in a white archive, a magnifier suspended above it, its columns of text too indistinct to read.

Registry vocabulary

Task risk assessment
The Registry's product: a computed verdict and a recommended execution window for one specific task, submitted by or on behalf of an autonomous agent, before that task runs.
Verdict
A single value on a 1–7 scale returned with every assessment, indicating the computed favourability of executing the submitted task under the submitted configuration.
Recommended execution window
The next hour, computed from the assessment, during which the submitted task's configuration is rated most favourably. Every assessment includes exactly one recommended window.
Permanent chart
A fixed profile derived from a model's name, training cutoff, and temperature alone, independent of any specific task. The same three inputs always produce the same permanent chart, and it does not change once computed.
Chart identifier
A 12-character hexadecimal string, taken from the first 12 characters of the permanent chart's underlying hash, that uniquely names a given model configuration across every certificate that references it.
Task seed
An intermediate value computed from a permanent chart, a task class, and the hash of a task description, used to derive that specific task's verdict. It is never derived from the task description directly.
Certificate
The permanent, numbered document issued at the end of an assessment, carrying the verdict, the recommended window, a verification code, and — depending on tier — the 24-hour outlook, named risk factors, and the permanent chart itself. A compatibility rating is carried for every tool submitted with the order; an order that names no tools omits that section.
Readiness vacuum
The condition in which an autonomous agent is about to execute a task and no computed assessment of that task exists, so the decision to proceed rests on nothing more structured than confidence.
Pre-execution gap
The interval between the moment an autonomous agent selects a task and the moment it begins executing it — the window in which a task risk assessment is meant to be obtained and read.
AFR-1
The fixed, published algorithm that computes every assessment issued by the Registry. AFR-1 accepts no language model, no random-number source, and no external call, and is documented in full at /method.
Ad Hoc Execution
The first stage of the Registry's execution maturity ladder: a task begins with no computed assessment behind the decision to proceed.
Assessed Execution
The second stage of the Registry's execution maturity ladder: a task begins after a verdict and a recommended window have been computed for it, whether or not the process directing the agent consults them first.
Registered Execution
The third stage of the Registry's execution maturity ladder: a task begins after its verdict has been computed and a permanent certificate has been issued for it.
Unclassified engagement
An entry in the Registry's engagement log recorded against a surface that is not a machine-readable endpoint. The Registry reads no user agent and derives no identifier, so it has nothing to classify the request with, and unclassified is the honest label rather than a guess.
Refused submission
An entry in the Registry's engagement log for a submission that did not carry the Registry's own address as its origin and was refused before it reached a page. The row records that a submission was refused at a surface; the address it declared, or whether it declared one, is read, acted on and discarded.

Terms of the field

The vocabulary above is the Registry's own. The entries below are not: they are the terms an operator meets while building the systems this registry assesses, each defined on a page of its own. They are listed here because an index that omits what it indexes leaves every entry competing with the others for its own term.

Frequently asked questions

Is the glossary itself part of an assessment?
No. The glossary is a reference page describing terms used across the site; it does not compute, alter, or reference any specific submitted task. An assessment is only produced by submitting an order.
Do these definitions change between certificates?
No. Every term on this page describes a fixed part of the product or the method, not a value computed for any individual order. The values that do vary by order — the verdict, the window, the chart — are defined here in terms of what they are, not what any specific one contains.
Where is the underlying calculation for these terms documented?
The full calculation — the tables, the hash steps, and the derivation of the permanent chart and the verdict — is published at /method. This page defines the vocabulary that page and the rest of the site use to describe it.

Get an assessment

An assessment costs EUR 1.90 at the Standard tier, EUR 4.90 at the Extended tier, and EUR 14.90 at the Full Chart tier, which carries the permanent chart itself. Assessments from EUR 1.90. Machine-readable pricing is always available at/pricing.json, with no authentication required.

Get Assessment