refactor(schema): split lead-scoring schema into schemes/lead_scoring/ [LTV-Pg.2]

.agent-plan.md

@@ -48,8 +48,9 @@ snapshots/relational_snapshot_safe/tasks moved + relational.py split so the
 shared write_relational_tables stays in the envelope) merged (#110). `LTV-Pg`
 (schema reorg) split: Pg.1 (scaffold `schemes/lifecycle/` — lifecycle rows +
 registries + LIFECYCLE_CONSTRAINTS moved there, stub LifecycleScheme registered)
-opened as **#111**; Pg.2 (split the lead-scoring schema out of shared `schema/`)
-pending. All M2 moves byte-identical. Sibling `leadforge-datasets-private`
+merged (#111); Pg.2 (split lead-scoring schema: entity rows/ALL_ROW_TYPES/
+ALL_CONSTRAINTS/LEAD_SNAPSHOT_FEATURES/CONVERTED_WITHIN_90_DAYS moved to
+`schemes/lead_scoring/`; shared primitives stay in `schema/`) opened as **#112**. All M2 moves byte-identical. Sibling `leadforge-datasets-private`
 consumes bundle files, not internals — no lockstep update needed (heads-up
 issue #8). Next: `LTV-Pg.2`, then `LTV-Pc` (pLTV feature/task specs, authored in
 `schemes/lifecycle/`), then `LTV-M3` (lifecycle population).

CLAUDE.md

@@ -156,7 +156,9 @@ leadforge/
   cli/            main.py, commands/{generate,list_recipes,inspect,validate}.py
   core/           rng.py, ids.py, time.py, enums.py, models.py, exceptions.py, ...
   narrative/      spec.py, company.py, product.py, personas.py, market.py, funnel.py, dataset_card.py
-  schema/         entities.py, relationships.py, events.py, features.py, tasks.py, dictionaries.py
+  schema/         entities.py (EntityRowProtocol, make_empty_dataframe, AccountRow — shared primitives),
+                  features.py (FeatureSpec), relationships.py (FKConstraint, validate_fk),
+                  tasks.py (SplitSpec, TaskManifest), dictionaries.py, tables.py
   schemes/        base.py (GenerationScheme protocol + SCHEME_REGISTRY);
                   lead_scoring/ — the lead-scoring scheme: __init__.py (build_world/
                   write_bundle) + simulation/, mechanisms/, structure/, render/

docs/ltv/roadmap.md

@@ -42,7 +42,7 @@ protocol + registry, with the package physically reorganized into
 |-----------|------------|-----|------------|
 | `LTV-M0` | Planning + design lock | `LTV-Pa` | #102, #103 (+ scheme reframe) |
 | `LTV-M1` | Lifecycle schema foundation | `LTV-Pb`, `LTV-Pc` | #104 (Pb) |
-| `LTV-M2` | Generation-scheme architecture + physical reorg | `LTV-Pd`, `LTV-Pe`, `LTV-Pf`, `LTV-Pg` | #107 (Pd), #108 (Pe), #109 (Pf.1), #110 (Pf.2), #111 (Pg.1) |
+| `LTV-M2` | Generation-scheme architecture + physical reorg | `LTV-Pd`, `LTV-Pe`, `LTV-Pf`, `LTV-Pg` | #107 (Pd), #108 (Pe), #109 (Pf.1), #110 (Pf.2), #111 (Pg.1), #112 (Pg.2) |
 | `LTV-M3` | Customer population + lifecycle world | `LTV-Ph`, `LTV-Pi` | |
 | `LTV-M4` | Lifecycle simulation engine | `LTV-Pj`, `LTV-Pk` | |
 | `LTV-M5` | Customer snapshots + pLTV targets (both regimes) | `LTV-Pl`, `LTV-Pm` | |
@@ -142,7 +142,7 @@ Total: ~19 PRs across 9 milestones.
     Shared primitives (`EntityRowProtocol`, `_empty_df`, `AccountRow`,
     `FKConstraint`) stay in `schema/` and are imported. Byte-identical;
     full suite green. (**PR #111**)
-  - [ ] **`LTV-Pg.2`** — split the **lead-scoring** schema: move the
+  - [x] **`LTV-Pg.2`** — split the **lead-scoring** schema (**PR #112**): move the
     lead-scoring entity rows / `ALL_ROW_TYPES` / `ALL_CONSTRAINTS` /
     `LEAD_SNAPSHOT_FEATURES` / task specs into `schemes/lead_scoring/`, leaving
     only genuinely shared primitives in `schema/`. (The lifecycle `LTV-Pc`

leadforge/exposure/filters.py

@@ -6,7 +6,7 @@
 
 The per-feature redaction policy lives separately on
 :attr:`leadforge.schema.features.FeatureSpec.redact_in_modes` and is queried
-via :func:`leadforge.schema.features.redacted_columns_for`.  ``BundleFilter``
+via :func:`leadforge.schemes.lead_scoring.features.redacted_columns_for`.  ``BundleFilter``
 deliberately does *not* duplicate that information so that the writer and
 the validator both consult the same source of truth.
 

leadforge/narrative/dataset_card.py

@@ -12,7 +12,7 @@
 from collections import Counter
 from typing import TYPE_CHECKING
 
-from leadforge.schema.features import LEAD_SNAPSHOT_FEATURES, FeatureSpec
+from leadforge.schemes.lead_scoring.features import LEAD_SNAPSHOT_FEATURES, FeatureSpec
 
 if TYPE_CHECKING:
     from leadforge.core.models import WorldSpec

leadforge/schema/dictionaries.py

@@ -11,7 +11,8 @@
 
 import pandas as pd
 
-from leadforge.schema.features import LEAD_SNAPSHOT_FEATURES, FeatureSpec
+from leadforge.schema.features import FeatureSpec
+from leadforge.schemes.lead_scoring.features import LEAD_SNAPSHOT_FEATURES
 
 _COLUMNS = ("name", "dtype", "description", "category", "is_target", "leakage_risk")