C3 — trail_stop_atr, roll_to_break_even and scale_out are implemented, tested, and never called

[eaitbrahim]

PRD §3 C3. The cheapest real win in the milestone.

The finding

keel/execution/executor.py implements and unit-tests three exit primitives:

• trail_stop_atr — ATR-multiple trailing stop, ratchet-only (never widens, consistent with rail 9)
• roll_to_break_even — move the stop to entry once a threshold is cleared
• scale_out — partial exit

A repo-wide search finds zero callers anywhere in keel/agent.py or outside executor.py and its own tests. They are built, tested, and dormant.

This matters beyond the missing feature: reading executor.py in isolation overstates what keel actually does. Anyone auditing the module would reasonably conclude keel trails its stops. It does not.

Competitive context

Freqtrade's trailing stop (trailing_stop_positive, trailing_stop_positive_offset, trailing_only_offset_is_reached) and adjust_trade_position partial exits are core, heavily-used capabilities. Jesse exposes the same through update_position().

Why it may have been left dormant — resolve before wiring

Not obviously an oversight, and worth establishing which:

1. Rail 9 (no stop-loss widening) requires a proposed stop to be strictly tighter than the last recorded one. A ratchet-only trailing stop should satisfy that by construction — confirm it does.
2. Exits currently ride a native Coinbase trigger-bracket placed at entry (executor.py:1129-1154), constructed as a raw dict against the pre-port CoinbaseClient. Moving a stop means amending or replacing that bracket, and the broker port has no bracket/OCO order kind at all — so trailing may have been blocked on the port, not forgotten.
3. turtle_breakout's real exit is the Donchian channel, not the nominal target_rr=6. Trailing may be redundant or actively harmful there, while helping pullback_continuation and rsi_meanrev.

Scope

• Establish why they are unwired; if it is (2), that is a port issue and should be split out.
• Wire whichever are safe, per rule family rather than globally.
• Backtest the difference before and after — the production-faithful engine (The simulator fills at a price live never waits for: entries are market orders in production, resting orders in backtest #257/fix(strategy): fill entries at the next bar's open, as production's market orders do #258) makes that measurable.
• If any stays dormant deliberately, say so in the module docstring, so the next reader is not misled.