Fix two incorrect comments about template maneuver XMLIDs

Defect 1: Comments incorrectly claimed all 53 template maneuvers carry
XMLID="MANEUVER". FACT: template maneuvers carry NO XMLID — DISPLAY is
their sole identity. The XMLIDs all being "MANEUVER" applies only to
maneuvers written into HDC files, not the template itself. That's why the
template index is keyed by display.

Corrected comments at:
- kirby_cost/campaign/rules.py:147 (comment)
- kirby_cost/campaign/rules.py:155-158 (ValueError message text)
- kirby_cost/template/hdt_provider.py:720-722 (docstring)

Defect 2: Comment at rules.py:41-43 claimed nine fields were "observed
changing a real object". FACT: only six were probed on real character
loads (killing, does_body, does_damage, does_knockback, defense, target,
range, uses_end, level_cost, display, level_value, level_power,
level_multiplier, min_set, max_set, base_value). The other three
(all_cost, group_cost, sense_cost) were probed via apply_template on a
constructed SenseAdder, not a real character load.

Narrowed the claim and added caveat explaining the testing method for
those three fields.

All 53 tests still passing. Refusal-message assertions still hold.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Xof4ZUS5n6A5ibX3PXNYbs

Author: pdbethke

kirby_cost/campaign/rules.py

@@ -39,8 +39,9 @@
 #:     level_multiplier, min_set, max_set: read directly by `apply_template`
 #:     (base.py:398, :487-:509, :545, :551) and by `Adder.apply_template`.
 #:   * The nine below were the ones the first pass never probed. Each was
-#:     then forced through `CampaignRules` and observed changing a real
-#:     object, on 2026-08-25:
+#:     then forced through `CampaignRules` and observed changing a loaded
+#:     object, on 2026-08-25 (six via real character loads; three via
+#:     constructed objects, noted below):
 #:   * types      -> forcing ("PROBE",) on RKA gave Ravel's RKA `_types
 #:                   == ["PROBE"]` (base.py, the `tmpl.types` loop).
 #:   * attributes -> forcing SHOWOPTIONONLY="Yes" on PENETRATING flipped
@@ -53,8 +54,11 @@
 #:   * base_value -> forcing 42.0 on STR moved Ravel's STR `base_level`
 #:                   from 10.0 to 42.0 (hdc_loader.py:655).
 #:   * all_cost / group_cost / sense_cost -> forcing 55.0 on a sense-rate
-#:                   xmlid put 55.0 on all three of a SenseAdder
+#:                   xmlid put 55.0 on all three of a constructed SenseAdder
 #:                   (base.py, the sense-rate loop; hdc_loader.py:1615).
+#:                   (These three were probed via apply_template on a built
+#:                   object, not a real character load, because none of the
+#:                   authored characters carries a SenseAdder.)
 #:
 #: The one field that is neither wired nor listed as a rule is `class_name`,
 #: below: it has ZERO reads off a `TemplateData` anywhere in `kirby_cost/`.
@@ -144,20 +148,20 @@ def set(self, xmlid: str, field: str, value: Any) -> None:
                 f"a field to its class default is not a capability that "
                 f"exists today.)"
             )
-        # Every one of the 53 template maneuvers carries XMLID="MANEUVER", so
-        # the maneuver index is keyed by DISPLAY and `hdc_loader` routes them
-        # to `get_maneuver`, never to `get_template_data`. "MANEUVER" is
-        # nonetheless IN the flat index (as Basic Strike, first wins), so the
-        # existence check below would accept this rule and it would be 100%
-        # inert. Refused here instead, before the check that would pass.
+        # Template maneuvers carry no XMLID — DISPLAY is their sole identity.
+        # So `hdc_loader` routes them to `get_maneuver` (keyed by display), not
+        # to `get_template_data`. If allowed here, this rule would find nothing
+        # on the template side and do nothing. Moreover, patching here by xmlid
+        # would rewrite all 53 maneuvers at once (they all parse with the same
+        # XMLID="MANEUVER" in HDC). Refused here instead, before the check.
         if xmlid.upper() == "MANEUVER":
             raise ValueError(
-                "'MANEUVER' cannot be the subject of a campaign rule: all 53 "
-                "template maneuvers share XMLID=\"MANEUVER\" and are looked "
-                "up by their display name, so this rule would never be read "
-                "-- and applying it by xmlid would rewrite all 53 at once. "
-                "Overriding one maneuver needs a rule keyed by display, "
-                "which does not exist yet."
+                "'MANEUVER' cannot be the subject of a campaign rule: template "
+                "maneuvers carry no XMLID (DISPLAY is their identity), so this "
+                "rule would find nothing on the template side. Moreover, all 53 "
+                "HDC maneuvers are written XMLID=\"MANEUVER\", so patching by "
+                "xmlid would rewrite all 53 at once. Overriding one maneuver "
+                "needs a rule keyed by display, which does not exist yet."
             )
         if self._provider.get_template_data(xmlid) is None:
             raise ValueError(

kirby_cost/template/hdt_provider.py

@@ -717,12 +717,12 @@ def get_maneuver(self, display: str) -> Optional[TemplateData]:
         None means "custom maneuver": Java builds one from the HDC element
         alone when no template maneuver matches the display.
 
-        NOT campaign-patched, deliberately. All 53 template maneuvers carry
-        XMLID="MANEUVER" -- that is why this index is keyed by display at all
-        -- so patching here by xmlid would apply one MANEUVER rule to all 53
-        at once. `CampaignRules.set()` therefore REFUSES "MANEUVER" outright,
-        so there is no accepted rule for this door to drop. A rule keyed by
-        maneuver display is a real feature; it is not this one.
+        NOT campaign-patched, deliberately. Template maneuvers carry no XMLID
+        (DISPLAY is their sole identity), so this index is keyed by display.
+        `CampaignRules.set()` therefore REFUSES "MANEUVER" outright: the rule
+        would find nothing on the template side, and even if it did, all 53
+        HDC maneuvers parse as XMLID="MANEUVER", so patching would rewrite all
+        53 at once. A rule keyed by maneuver display is a real feature, not this.
         """
         return self._maneuvers.get(display)