docs: non-warhammer system drop policy (Middle-Earth default-dropped)

AGENTS.md

@@ -225,9 +225,32 @@ never re-fetch them again):
    (`CROSS_SYSTEM_MIN = 0.75`) - weaker guesses stay in `_unsorted`.
    Token-containment boosts exist for short names vs long catalog names
    ("telemon" → "Telemon Heavy Dreadnought"), guarded by a generic-token
-   stoplist. The catalog deliberately covers all GW systems (needed for
+   stoplist.
+
+   **Refined 2026-10-07 (dir re-homing + weak tier + non-warhammer drop):**
+   - **Deepest-matching-dir wins**: ancestor directory names (inside the
+     archive in provenance mode, real dirs in manual trees) compete with
+     the pack/file name; the best accepted match re-homes the whole
+     subtree. Junk dir names (files/images/HEADS/pure numbers/lone generic
+     words) are never candidates. Fixes multi-unit sculptor packs (one
+     "Retro Elvez" archive ≈ 20 Eldar units no longer all land in one
+     keyword class).
+   - **Weak tier**: when nothing reaches the threshold, a hit >= `WEAK_MIN`
+     (0.50) is accepted but flagged low confidence - `[w]` in the plan,
+     `confidence: low` in the audit, listed in
+     `<library>/sorter_low_confidence.txt`. Weak and tokenless adoptions
+     require a shared non-generic token; tokenless normal adoption needs
+     +0.08 (`TOKENLESS_BONUS`); cross-system adoption additionally needs
+     token containment (`_token_contained`), because char-ratio lies
+     ("aeldari rangers" vs "dwarf rangers" = 0.786).
+   - **Non-warhammer systems dropped**: `DROPPED_SYSTEMS` (default
+     `{Middle-Earth}`) - catalog entries filtered at load, keyword hits
+     count as unclassified; files stay in `_unsorted`. `--drop-system`/
+     `--keep-system` adjust. The catalog snapshot keeps all systems.
+
+   The catalog deliberately covers all GW systems (needed for
    catalog-agnostic use) - wrong-system placement is prevented by matching
-   discipline, not by shrinking the catalog.
+   discipline and the drop filter, not by shrinking the catalog.
 
    **Catalog source: the GW webshop via Algolia (`dlphelper/gwshop.py`,
    working, verified 2026-10-04 — how it was obtained, do not re-derive):**

docs/SORTING.md

@@ -140,6 +140,14 @@ already-present duplicates, 0 copied):
   truncated/HTML dead-link downloads are listed in
   `<extracted>/.broken/RETRY_URLS.txt` (all 0807.st; re-run
   `python3 dlp.py <channel> <out> --urls-file ...` for them).
+- **Non-warhammer systems are dropped** (2026-10-07, user decision:
+  "drop anything that isn't really warhammer"). The sorter default
+  (`DROPPED_SYSTEMS` in `dlphelper/sorter.py`) is `{Middle-Earth}`:
+  catalog entries of dropped systems are filtered at load and keyword
+  hits for them count as unclassified, so nothing is ever placed into
+  those sections - the files stay in `_unsorted`. `--drop-system NAME`
+  drops more, `--keep-system NAME` re-enables one. The committed catalog
+  snapshot itself is untouched.
 
 ## Catalog file (`data/library_catalog.json`)