DEM-PRODUCER-PLAN (core): delete the rrMultiplePlanned synthesis; generalize the composer

[Gio2050]

Wave 1 of 4 — DEM-PRODUCER-PLAN, step 1 of 3. Merge this first, then afi-config dem/plan-config, then afi-reactor dem/plan-reactor.

The accepted row and gate (DEM-GOV §9, quoted before building)

DEM-PRODUCER-PLAN | D-DEM-4(2), D-DEM-5(1)(6): rrMultiplePlanned carried from the submitted trade plan through cpjMapper as a provider fact, against the governed trade-plan contract; deletion of the synthesis | D-DEM-5(6) satisfied — the submitted plan is verified against observed market data by the producer that emits it, and refused when it cannot be verified, proven by test; the value is a provider fact, never synthesized from an analyst input — the analyst/provider role split proven by test; absence fails closed under D-DEM-5(2) unless declared optional within D-DEM-5(4)(b); the risk/structure collinearity of Evidence §3 recorded as resolved or recorded as residual, per D-DEM-5(5); movement itemized; implementationVersion bumps per D-DEM-7(4)

Authorization: your instruction of 2026-08-25, recorded verbatim in the DEM-GOV Status line and §9 (afi-governance #55, merge 7274627); determinations D-2/D-4/D-5 in #56.

What this PR does

The planned R:R stops being invented. rrMultiplePlanned was synthesized from an analyst input (pulledBackIntoSweetSpot && !brokeEmaWithBody ? 2 : 1) — the exact role-split violation D-DEM-4(2) names. That line is deleted. The value now comes from the submitted trade plan, verified by the technical lane against the candles it fetched, and bound by the registered mapping.

• froggy.enrichment_adapter.ts — the synthesis is gone; the builder returns FroggyAdapterOutput (no rrMultiplePlanned); the view type gains technical.plan (the lane's verified plan facts, read only through the mapping).
• froggy.residual_builder.ts — the composer is now mapping-version-agnostic and fail-closed: fragment ∪ residual must partition the ten scorer-input fields exactly (nothing missing, nothing supplied twice) and every value must lie in its declared domain. The domain table is predicate-valued on purpose, so no scorer-input field is ever assigned a literal in this module — the coming GUARD slot has nothing to flag here.
• froggy.trend_pullback_v1.ts — scoreFroggyTrendPullbackFromEnriched is deleted: the adapter no longer produces a full input, so that convenience path could only lie.
• Tests — the registered 1.1.0 mapping as shared test data (inline, drift-checked against the sibling registry file when present); the byte-equivalence suite now proves compose(fragment, residual) equals the adapter's output plus the mapping-bound R:R, with the fired default recorded when the producer emitted none; composer refusal cases for missing / extra / dual-source / out-of-domain.

Verification

afi-core 312/312 green; tsc --noEmit green. No scoring-law value moves (weights, clamps, R:R bands, riskBucket map untouched).

🤖 Generated with Claude Code

[kilo-code-bot]

Kilo Code Review could not run — your account is out of credits.

Add credits or switch to a free model to enable reviews on this change.