\label{def:physical-e2-agent}
A \textbf{physical $\Elayer{2}$ agent} is an $\Elayer{2}$
computational agent $(C, D, \varphi)$
(Definition~\ref{def:e2-computational-agent})
whose carrier is $\Elayer{1}$-admissible:
\begin{enumerate}
\item[(i)] Code $C$, decoder $D$, and all intermediate states
      $C_n$ are $\Elayer{1}$-objects: finite spatial extent,
      bounded energy, sector addresses in the $4{+}1$
      decomposition.
\item[(ii)] The execution map $\varphi \colon C_n \to C_{n+1}$
      respects $\Elayer{1}$ causal structure: each step
      involves only causally connected regions
      (Remark~\ref{rem:light-cone-consistency}, Ch.~77).
\item[(iii)] Operational closure holds within the $\Elayer{1}$
      carrier: the cycle does not require external resources
      that grow without bound.
\end{enumerate}
Unlike the $\tau$-Tower Machine (which is $\tau$-native),
a physical $\Elayer{2}$ agent inherits its constraints from
the $\Elayer{1}$ physics of its carrier.