feat(loader): REQUIRED template adders attach as HD's getAssignedAdders() does

GenericObject.getAssignedAdders() clones every REQUIRED available adder not
already assigned and marks it selected -- generic, on READ. HD-written files
state these adders, so the corpus never exercised the attach; a hand-written
element without them (the sink's Damage Over Time) got a different value
(+1 vs HD's +3 1/4: base 1 + INCREMENTS first option 0.25 + TIMEBETWEEN
first option 2.0). The engine now materialises them at the end of
_build_modifier -- AFTER the element's own adders (attaching before them
double-counted 24 corpus objects) -- resolving the SAME template variant the
modifier application used (MINDLINK's nested INVISIBLE has no required
adders; the global one does), priced at the first option, alias wired.

Sink cost ledger 10 -> 8; the last prototype cell's reason now carries the
loaded-vs-prototype cross-evidence. Suite 1,728; parity and display intact.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PrVDXgpDfQpzEGErZTLjEs

Author: pdbethke

kirby_cost/io/hdc_loader.py

@@ -1872,6 +1872,44 @@ def _build_modifier(self, elem, parent: GenericObject) -> Optional[Modifier]:
                                                option_id=adder_opt if adder_opt else None)
                 mod.assigned_adders.append(adder)
 
+        # HD attaches every REQUIRED available adder the moment the list is
+        # read -- GenericObject.getAssignedAdders() clones any required
+        # available adder not already assigned and marks it selected -- and
+        # the attached adder prices at its FIRST option (the sink's Damage
+        # Over Time: base 1 + INCREMENTS "2" 0.25 + TIMEBETWEEN SEGMENT 2.0 =
+        # HD's +3 1/4, both first options). HD-written files always state
+        # these adders explicitly, so the corpus never exercises the attach;
+        # a hand-written element without them must still get HD's answer.
+        # Runs AFTER the element's own ADDER children are parsed -- attaching
+        # before them duplicated every stated required adder (24 corpus
+        # failures, caught 2026-08-31).
+        # Resolve the SAME template variant the modifier application used --
+        # a nested definition (MINDLINK's own INVISIBLE has NO required
+        # adders) beats the global one (whose EFFECTSTARGET/EFFECTSOTHER are
+        # REQUIRED); attaching from the global variant garbled nine MindLinks.
+        tmpl_late = None
+        provider = self._provider_in_use
+        if parent is not None and hasattr(provider, "get_nested_modifier"):
+            tmpl_late = provider.get_nested_modifier((parent.xmlid or "").upper(), xmlid)
+        if tmpl_late is None:
+            tmpl_late = self._get_template_data(xmlid)
+        for adder_xmlid, adder_tmpl in (getattr(tmpl_late, "adders", None) or {}).items():
+            if not getattr(adder_tmpl, "required", False):
+                continue
+            if GenericObject.find_object_by_id(mod.assigned_adders, adder_xmlid) is not None:
+                continue
+            adder = Adder()
+            adder.xmlid = adder_xmlid
+            adder._required = True
+            adder._selected = True
+            adder._alias = getattr(adder_tmpl, "alias", "") or getattr(adder_tmpl, "display", "")
+            adder.apply_adder_template(adder_tmpl)
+            opts = getattr(adder_tmpl, "options", None) or {}
+            if opts and not adder.base_cost:
+                first_id, first = next(iter(opts.items()))
+                adder.base_cost = getattr(first, "base_cost", 0.0) or 0.0
+                adder.option_id = first_id
+            mod.assigned_adders.append(adder)
         return mod
 
     def _build_adder(self, elem) -> Optional[Adder]:

tests/fixtures/included_known_gaps.json

@@ -2,6 +2,6 @@
   "_comment": "SHRINK-ONLY. Cells where the engine's included() disagrees with Hero Designer's, with what each side said. Written once as the 2026-08-29 baseline before any fix; every later commit may only remove entries. The target is empty.",
   "_baseline": "3770 of 8395 cells disagree (2026-08-29); now 1 remain (2026-08-30)",
   "gaps": {
-    "REDUCEDEND-on-SHAPESHIFT": "HD allows; unexplained. The visible Java path (ReducedEND.java:200-210) refuses on getEndUsage()==0, HD's own echo for this prototype says end_usage=0 and active_cost=0.0, and neither Shapeshift.java nor SenseAffectingPower.java overrides getEndUsage(). Suspect: the o.clone()/setParent(null) at ReducedEND.java:130-132 changing what the clone's getEndUsage() sees. Needs a live-HD trace; the last prototype cell (2026-08-30)."
+    "REDUCEDEND-on-SHAPESHIFT": "HD allows; mechanism unlocated after a full read of ReducedEND.java:124-210 -- the visible path refuses on getEndUsage()==0, HD's own echo for this prototype says end_usage=0/active_cost=0.0, no getEndUsage/getActiveCost override exists in Shapeshift/SenseAffectingPower, and clone() re-selects the same (null) option. Cross-evidence: on the LOADED sink Shape Shift (stateful fixture, object 20260830000064) HD reports end_usage=1/active=8.0 and allows -- consistent with its rule -- and the engine agrees there. The divergence is confined to HD's PROTOTYPE tier. Last prototype cell; needs a live-HD debugger trace (2026-08-31)."
   }
 }

tests/fixtures/validation_sink_cost_gaps.json

@@ -1,14 +1,12 @@
 {
   "_comment": "Shrink-only ledger for tests/test_validation_sink_costs.py. Regenerate the fixture: venv/bin/python -c 'from tests.validation_sink import write; write(\"/tmp/ValidationSink.hdc\")' ; (cd ../kirby-hd-oracle && ./hd6cli.sh /tmp/ValidationSink.hdc) > tests/fixtures/authored/ValidationSink.json. Seeded 2026-08-30: the corpus is HD-written and cannot see these; the hand-written sink can.",
-  "_baseline": "23 sink cost values disagree with HD (2026-08-30); option-cost fix 2026-08-30: 13 closed, now 10 remain; 5E-fallback ruling + MULTIPLESFX state 2026-08-30: 5 closed, 5 new, now 10",
+  "_baseline": "23 sink cost values disagree with HD (2026-08-30); option-cost fix 2026-08-30: 13 closed, now 10 remain; 5E-fallback ruling + MULTIPLESFX state 2026-08-30: 5 closed, 5 new, now 10; REQUIRED-adder attach 2026-08-31: 2 closed (DoT), now 8",
   "gaps": {
     "20260830000055:real_cost": "framework: slot/pool cost rule differs on this hand-written shape (HD zeroes or redistributes where the engine costs the slot standalone); needs its own read of List/Multipower/VPP cost paths -- HD 27.0 engine 40.0 (ENERGYBLAST 'VPP Blast Slot')",
     "20260830000057:real_cost": "framework: slot/pool cost rule differs on this hand-written shape (HD zeroes or redistributes where the engine costs the slot standalone); needs its own read of List/Multipower/VPP cost paths -- HD 7.0 engine 11.0 (FORCEFIELD 'VPP Force Field Slot')",
     "20260830000060:active_cost": "carrier math: Absorption's identical AOE closed with the option-cost fix, so the residual is the CARRIER's own cost rule on this shape (Images' +4m pricing / NakedModifier's base), not the AOE value -- HD 15.0 engine 10 (NAKEDMODIFIER 'Naked Advantage')",
     "20260830000060:real_cost": "carrier math: Absorption's identical AOE closed with the option-cost fix, so the residual is the CARRIER's own cost rule on this shape (Images' +4m pricing / NakedModifier's base), not the AOE value -- HD 15.0 engine 10 (NAKEDMODIFIER 'Naked Advantage')",
     "20260830000061:real_cost": "CustomPower real-cost rounding on a hand-written shape -- HD 0.0 engine 1.0 (CUSTOMPOWER 'Custom Power With END')",
-    "20260830000074:active_cost": "modifier value: HD's DoT prototype carries template-REQUIRED adders (damage increments, frequency) the sink never stated; engine sums neither -- HD 170.0 engine 80 (ENERGYBLAST 'Blast Damage Over Time')",
-    "20260830000074:real_cost": "modifier value: HD's DoT prototype carries template-REQUIRED adders (damage increments, frequency) the sink never stated; engine sums neither -- HD 170.0 engine 80.0 (ENERGYBLAST 'Blast Damage Over Time')",
     "20260830000124:total_cost": "carrier math: Absorption's identical AOE closed with the option-cost fix, so the residual is the CARRIER's own cost rule on this shape (Images' +4m pricing), not the AOE value -- HD 10.0 engine 16.0 (IMAGES 'Images')",
     "20260830000124:active_cost": "carrier math: Absorption's identical AOE closed with the option-cost fix, so the residual is the CARRIER's own cost rule on this shape (Images' +4m pricing), not the AOE value -- HD 12.0 engine 20 (IMAGES 'Images')",
     "20260830000124:real_cost": "carrier math: Absorption's identical AOE closed with the option-cost fix, so the residual is the CARRIER's own cost rule on this shape (Images' +4m pricing), not the AOE value -- HD 12.0 engine 20.0 (IMAGES 'Images')"