AoA Agent Registry: a dated snapshot of the running agents of the Assessment on the AI framework, 4 September 2026

This is a dated snapshot, not a maintained catalogue. It records which agents of the Assessment on the AI (AoA) framework were running on the CAFA platform on 4 September 2026, what each was instructed to do, where a human decision was required, and what kind of guarantee each output carries. Fourteen agents are documented: a gateway agent and thirteen functional agents covering curriculum ontology construction (COREA), lesson planning (ALPHA), item engineering (K-SALT), blueprint assembly (BASIC), adaptive delivery (AQUA), rubric generation (GARAM), developmental evaluation (GRADE), reverse-engineered passage construction (REPAIR), inventory-driven adaptive learning over a standards-indexed template pool spanning the U.S. Common Core and the Korean national mathematics curriculum (CIA), a dual-agent survey pair for perception and feedback data (ASSISI), and a dual-agent essay assessment pair whose configuration agent is fully symbolic and involves no language model (BEST). For each agent the registry states its identifier and creation date, its function, the point at which a human decision is required, and the kind of guarantee its output carries, using a four-part distinction: machine-verified, measured but approximate, provisional, and recorded only. The platform is under active development and its agents change. This snapshot will not be updated as the platform changes; the claim it supports is that on the stated date these agents existed and operated as described, which does not become false when the platform moves on. A running platform closes the question of whether an architecture can be built. It leaves the question of performance entirely open. No usage data, output quality data, or learner outcomes are reported here, and the human checkpoints documented are specified rather than measured: the registry states where a human decision is required, not how often it was exercised.