Activation generator and ledger are hardcoded in the loop; ledger.py and milp_bidding.py import upward from core.simulator

[dim-sous]

What is wrong

Two more concrete types the loop constructs by hand, and two imports that point the wrong way.

Hardcoded in the loop:

• core/simulator/core.py:45 imports and :84 constructs ActivationSignalGenerator(reg_p, tp.dt_pi). A different regulation signal model — recorded FCR data, a different spectral shape, a deterministic stress profile — is a core edit.
• core/simulator/core.py:32 imports compute_ledger and calls it at the end of the run. A different scoring — say, a settlement rule for another market — is a core edit.

Upward imports (leaf → trunk):

• core/accounting/ledger.py:29 — from core.simulator.traces import SimTraces
• core/planners/milp_bidding.py:1035 — from core.simulator.synthetic_day import make_synthetic_day (inside a function body, which is how these hide)

Nothing forbids either today because no boundary is declared. Once #1 lands the ledger is devkit and may import the simulator freely — but a planner importing the synthetic day is a product module reaching into the harness and becomes a hard violation of the one-way rule.

Done when

• The activation signal is a component the loop receives, behind a small Protocol, chosen at composition.
• Scoring is a component the loop receives (or is applied after the loop returns traces), not a call baked into it.
• milp_bidding.py does not import the synthetic day; whatever it needed from it moves to where a planner may look.
• The one-way test from The repo has no ships / never-ships boundary: plant, prices, ledger and the loop sit beside the controller in core/ #1 is green.
• Proof: traces and ledger output bit-identical.

Phase 1 of the platform epic. Depends on #1.