Agency Envelope Definition

The problem

The architecture requires distinguishing "person is in a state where capacity for self-directed action is degraded" from "person is making choices I wouldn't make." The neurobiological floor (PFC offline, reflexive mode) provides one boundary. But the space between "clearly degraded" and "clearly fine" is large and normatively contested.

Health-as-boundary-constraint (from A0) addresses part of this: health violations are the clearest signal the system is moving in the wrong direction. But the boundary between "degraded capacity" and "unusual but autonomous choice" is not fully specified.

Why this matters

If the envelope is drawn too narrowly, the system intervenes on normal variation and autonomous choice (paternalism). If drawn too broadly, the system misses genuine degradation. The boundary definition is itself a governance decision -- and critics will probe here, because the line between "maintaining conditions for autonomy" and "imposing a normative standard of functioning" is inherently contestable.

Partial resolution

The state-conditioned gating table and the U (uncertain) regime provide conservative defaults: when the boundary is unclear, default to null. But this defers rather than resolves the definitional question.