From 1e35ab947b5bc49395f11c2d199c72d5212b4ee4 Mon Sep 17 00:00:00 2001 From: DKChan <524912269@qq.com> Date: Mon, 2 Nov 2026 01:58:22 +0800 Subject: [PATCH] =?UTF-8?q?iter48-=E5=8D=A11=EF=BC=9A=E5=8D=A1=E2=91=A0(?= =?UTF-8?q?=E7=8A=B6=E6=80=81=E5=90=88=E5=B9=B6=E7=B4=AF=E5=8A=A0=E5=88=B6?= =?UTF-8?q?=E6=89=B9):=20=E5=85=AD=E8=A1=8Cstack=5Fpolicy=20refresh?= =?UTF-8?q?=E2=86=92accumulate+schema=E8=AF=8D=E8=A1=A8=E6=8D=A2=E4=BB=A3+?= =?UTF-8?q?=E5=AE=B9=E5=99=A8=E7=B4=AF=E5=8A=A0=E8=87=82(=5Fapply=5Faccumu?= =?UTF-8?q?late=E5=8F=AA=E5=8A=A0=E6=97=B6=E9=95=BF,=E5=B1=82=E6=95=B0?= =?UTF-8?q?=E5=BF=AB=E7=85=A7=E4=B8=8D=E5=8F=98)+roll=5Fmultiplier?= =?UTF-8?q?=E4=BA=94=E6=AE=B5=E5=BC=8F=E6=8E=A2=E9=92=88(V1/V2/V3)+table?= =?UTF-8?q?=5Fstructure=5Fguard=20V4+damage=5Fcontroller=E6=A0=87=E7=AD=BE?= =?UTF-8?q?=E6=8D=A2=E4=BB=A3+=E7=94=B2=E6=A1=88=E8=A3=81=E5=86=B3?= =?UTF-8?q?=E9=9A=8F=E5=8A=A8(2026-11-02=E7=94=A8=E6=88=B7=E8=A3=81):=20?= =?UTF-8?q?=5Fineffective=5Frow=E8=BD=BD=E8=8D=B7deferred=E8=87=AA?= =?UTF-8?q?=E6=8F=8F=E8=BF=B0=E9=94=AE(=E9=9D=9E=E8=A1=8C=E4=B8=BA?= =?UTF-8?q?=E6=94=B9=E5=8A=A8)+round=5Fflow=20=5Frow=5Fphase=E5=88=86?= =?UTF-8?q?=E7=B1=BB=E8=A1=A5deferred=E2=86=92=E6=88=98=E6=96=97=E6=AE=B5?= =?UTF-8?q?=E2=80=94=E2=80=94=E7=9B=B8=E4=BD=8D=E5=8D=95=E8=B0=83=E6=96=AD?= =?UTF-8?q?=E8=A8=80=E5=BD=92=E7=BB=BF,EV=E4=B8=89=E9=94=9A0.9148/1.0012/1?= =?UTF-8?q?.1948=E9=80=90=E4=BD=8D=E4=B8=8D=E5=8F=98?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- game/core/damage_controller.gd | 2 +- game/core/status_container.gd | 10 ++--- game/data/schemas/status_effect_table.gd | 2 +- game/data/tables/status_effects.tres | 12 +++--- .../damage_controller_test.gd | 2 +- .../battle_settlement/roll_multiplier_test.gd | 43 +++++++++++++------ .../round_flow_skill_pool_test.gd | 3 +- .../data_config/table_structure_guard_test.gd | 4 +- .../需求卡片-状态合并累加制批-2026-11-01.md | 6 +++ 9 files changed, 55 insertions(+), 29 deletions(-) diff --git a/game/core/damage_controller.gd b/game/core/damage_controller.gd index a27469fb..585eb080 100644 --- a/game/core/damage_controller.gd +++ b/game/core/damage_controller.gd @@ -688,4 +688,4 @@ func _target_for(rule: String) -> Variant: func _ineffective_row(t: Dictionary, reason: String) -> void: emit_row_cb.call({"type": "reward_ineffective", "source_class": String(t.source_class), "origin_cell": t.origin_cell, "highlight": String(t.highlight)}, int(t.round_index), - "%s 未生效（%s）" % [String(t.display), reason], 0, {"batch_key": String(t.context.get("batch_key", ""))}) + "%s 未生效（%s）" % [String(t.display), reason], 0, {"batch_key": String(t.context.get("batch_key", "")), "deferred": true}) diff --git a/game/core/status_container.gd b/game/core/status_container.gd index 52def7f8..391f2bd0 100644 --- a/game/core/status_container.gd +++ b/game/core/status_container.gd @@ -41,8 +41,8 @@ func apply_effect(effect_id: String, source: Dictionary, duration_override: int "independent": _instances.append(_new_instance(effect_id, 1, dur, source, snap)) return true - "refresh": - return _apply_refresh(effect_id, dur, snap, source) + "accumulate": + return _apply_accumulate(effect_id, dur, snap, source) "stack": return _apply_stack(effect_id, dur, snap, source, max_stack) push_warning("[增益状态] 未知 stack_policy") @@ -57,11 +57,11 @@ func _grant_snapshot(row: Dictionary, source: Dictionary) -> Dictionary: return {"m": snap_m, "unit": String(row.unit), "display": String(row.display_name)} -## refresh 臂：命中改 R 未命中追加（R ← D 取实例快照 D——同表值；L 不变；空刷新恒 true） -func _apply_refresh(effect_id: String, dur: int, snap: Dictionary, source: Dictionary) -> bool: +## 累加臂：命中加时长未命中追加（R += D 取本次授予生效时长——覆盖值 ≥1 按覆盖值；L 不变；快照沿首授；每次授予必有剩余次数变化） +func _apply_accumulate(effect_id: String, dur: int, snap: Dictionary, source: Dictionary) -> bool: for inst in _instances: if inst.effect_id == effect_id: - inst.remaining = dur + inst.remaining = int(inst.remaining) + dur return true _instances.append(_new_instance(effect_id, 1, dur, source, snap)) return true diff --git a/game/data/schemas/status_effect_table.gd b/game/data/schemas/status_effect_table.gd index 4e1568cb..0c15dffe 100644 --- a/game/data/schemas/status_effect_table.gd +++ b/game/data/schemas/status_effect_table.gd @@ -8,7 +8,7 @@ extends Resource const TARGET_SIDES := ["hero", "enemy"] const UNITS := ["percent", "flat", "ac"] # ac——AC 阈值轴效果（ac_up），非池百分比非定点值 -const STACK_POLICIES := ["stack", "refresh", "independent"] +const STACK_POLICIES := ["stack", "accumulate", "independent"] func validate() -> Array: diff --git a/game/data/tables/status_effects.tres b/game/data/tables/status_effects.tres index 91f46885..7833dfdd 100644 --- a/game/data/tables/status_effects.tres +++ b/game/data/tables/status_effects.tres @@ -11,7 +11,7 @@ rows = [{ "magnitude": 50, "unit": "percent", "duration_settlements": 1, -"stack_policy": "refresh", +"stack_policy": "accumulate", "max_stack": 2 }, { "effect_id": "ac_up", @@ -20,7 +20,7 @@ rows = [{ "magnitude": 1, "unit": "ac", "duration_settlements": 1, -"stack_policy": "refresh", +"stack_policy": "accumulate", "max_stack": 2 }, { "effect_id": "weaken", @@ -29,7 +29,7 @@ rows = [{ "magnitude": -65, "unit": "percent", "duration_settlements": 2, -"stack_policy": "refresh", +"stack_policy": "accumulate", "max_stack": 16 }, { "effect_id": "enrage", @@ -65,7 +65,7 @@ rows = [{ "magnitude": 100, "unit": "percent", "duration_settlements": 2, -"stack_policy": "refresh", +"stack_policy": "accumulate", "max_stack": 1 }, { "effect_id": "fortify", @@ -74,7 +74,7 @@ rows = [{ "magnitude": -50, "unit": "percent", "duration_settlements": 2, -"stack_policy": "refresh", +"stack_policy": "accumulate", "max_stack": 1 }, { "effect_id": "defensive_stance", @@ -83,7 +83,7 @@ rows = [{ "magnitude": -25, "unit": "percent", "duration_settlements": 2, -"stack_policy": "refresh", +"stack_policy": "accumulate", "max_stack": 1 }, { "effect_id": "alert", diff --git a/game/tests/unit/battle_settlement/damage_controller_test.gd b/game/tests/unit/battle_settlement/damage_controller_test.gd index f3b343a5..b42bde9b 100644 --- a/game/tests/unit/battle_settlement/damage_controller_test.gd +++ b/game/tests/unit/battle_settlement/damage_controller_test.gd @@ -373,7 +373,7 @@ func _test_pool_clamp(db) -> void: var expected1: int = clampi(int(pool_eff.magnitude), lo, hi) Kit.check("池值 = 单实例 L×m 加算后 clamp", first_ok and h.ctrl.clamped_pool("hero_offense") == expected1) var second_ok: bool = h.effects.apply_effect(String(pool_eff.effect_id), src) - Kit.check("refresh 政策重复授予受理且层数恒 1（叠层改续时裁决）——池值不变", + Kit.check("累加政策重复授予受理且层数恒 1（同名重复授予加时长）——池值不变", second_ok and h.ctrl.clamped_pool("hero_offense") == expected1) # ac 型效果不入 percent 池 var ac_eff2 := {} diff --git a/game/tests/unit/battle_settlement/roll_multiplier_test.gd b/game/tests/unit/battle_settlement/roll_multiplier_test.gd index ac4365af..df8438c4 100644 --- a/game/tests/unit/battle_settlement/roll_multiplier_test.gd +++ b/game/tests/unit/battle_settlement/roll_multiplier_test.gd @@ -154,13 +154,18 @@ func _smoke_armor_roll() -> void: int(crit_rows.get("roll", -1)) == int(_db.combat.settlement.damage_roll.crit_roll) and bool(crit_rows.get("crit", false))) Kit.check("nat20 → 普通分量 ×2（%d = 2×%d，暴击倍率只乘该分量，穿刺/魔法不吃）" % [int(crit_rows.get("normal", -1)), int(_db.combat.hero.atk)], int(crit_rows.get("normal", -1)) == 2 * int(_db.combat.hero.atk) and int(crit_rows.get("value", -1)) == 2 * int(_db.combat.hero.atk)) - # ③ 削弱续时（行为面）：授予→衰减一轮→再授予 → 层数保持 1、剩余结算数重置 - # （削弱为续时语义，无叠层通道） - var weaken_probe := _weaken_refresh_probe() - Kit.check("削弱重复授予 → 层数保持 1、剩余结算数重置 2（续时语义——结构性消饱和）", - int(weaken_probe.get("layers", -1)) == 1 and int(weaken_probe.get("remaining", -1)) == 2) + # ③ 削弱累加（行为面）：授予→衰减一轮→再授予 → 层数保持 1、剩余结算数累加本次授予时长 + # （削弱为累加语义，无叠层通道） + var weaken_probe := _weaken_accumulate_probe() + Kit.check("削弱重复授予 → 层数保持 1、剩余结算数累加至 3（1+2——快照沿首授）〔状态累加 V1〕", + int(weaken_probe.get("after_regrant", -1)) == 3 and int(weaken_probe.get("layers", -1)) == 1 \ + and int(weaken_probe.get("snap_magnitude", -9999)) == -65) Kit.check("削弱单层池读数 = 行值（enemy_output 池单层贡献——×2 叠满触底 ×0 不可达，免伤轮根治）", int(weaken_probe.get("pool", -999)) == int(_db.combat.status_effects.row_by_id("weaken").magnitude)) + Kit.check("削弱覆盖时长重授 → 按覆盖值累加至 6（3+3）〔状态累加 V3〕", + int(weaken_probe.get("after_override", -1)) == 6) + Kit.check("削弱同轮再授 → 再累加一段至 8（6+2）〔状态累加 V2〕", + int(weaken_probe.get("after_extra", -1)) == 8) # ④ AC11 折扣档分布（行为面）：ac_up 逐轮授给（时效 1）→ Boss 行攻击拍逐拍判定 # （ac_up 层数恒 1——局内可达 AC 上限 10+1=11，探针随行取值） var tiers := _ac11_tier_samples() @@ -215,10 +220,11 @@ func _first_hero_attack_rows(seed_val: int) -> Dictionary: return {} -# 削弱续时探测： -# 授予 → 衰减一轮 → 再授予，读 weaken 实例投影 {layers, remaining, pool}。 +# 削弱累加探测（五段式）： +# 授予(2) → 衰减一轮(1) → 重授(1+2=3) → 覆盖值重授(3+3=6) → 同轮再授(6+2=8)， +# 读 weaken 实例投影 {layers, pool, after_regrant, after_override, after_extra, snap_magnitude}。 # 直授通道（来源格真实坐标惯例——沿前例）；depth 2 板（D2 熊地精）仅作宿主，行为与敌种无关。 -func _weaken_refresh_probe() -> Dictionary: +func _weaken_accumulate_probe() -> Dictionary: var rng = load("res://core/rng_service.gd").new() rng.set_master_seed(20260916) var settlement = load("res://core/battle_orchestrator.gd").new(_db, rng) @@ -227,11 +233,24 @@ func _weaken_refresh_probe() -> Dictionary: var src: Dictionary = {"source_class": "mine_hit", "origin_cell": Vector2i(0, 0), "round_index": settlement.round_index} settlement._status_effects.apply_effect("weaken", src) settlement._status_effects.decay_round() # remaining 2 → 1（先兑付后衰减——只借衰减器，不结算） - settlement._status_effects.apply_effect("weaken", src) # 续时：remaining ← 2，层数不动 + settlement._status_effects.apply_effect("weaken", src) # 累加：remaining 1+2=3，层数不动 + var regrant := _weaken_inst(settlement) + settlement._status_effects.apply_effect("weaken", src, 3) # 覆盖时长按覆盖值累加：3+3=6 + var override_grant := _weaken_inst(settlement) + settlement._status_effects.apply_effect("weaken", src) # 同轮再授：6+2=8 + var extra := _weaken_inst(settlement) + if regrant.is_empty() or override_grant.is_empty() or extra.is_empty(): + return {} + # 走伤害控制器公有算术面 clamped_pool + return {"layers": int(extra.layers), "pool": int(settlement._damage.clamped_pool("enemy_output")), + "after_regrant": int(regrant.remaining_settlements), "after_override": int(override_grant.remaining_settlements), + "after_extra": int(extra.remaining_settlements), "snap_magnitude": int(regrant.snap_magnitude)} + + +func _weaken_inst(settlement: Object) -> Dictionary: for eff in settlement._status_effects.query_active("enemy"): if String(eff.effect_id) == "weaken": - # 走伤害控制器公有算术面 clamped_pool - return {"layers": int(eff.layers), "remaining": int(eff.remaining_settlements), "pool": int(settlement._damage.clamped_pool("enemy_output"))} + return eff return {} @@ -339,7 +358,7 @@ func _smoke_pool_wiring() -> void: # 构造停手轮首行主角攻击行：opts = {effects: Array}；hero_atk 覆盖 100 拉开量级 -# （floor 关系可精确断言）；效果直授沿 _weaken_refresh_probe 前例。 +# （floor 关系可精确断言）；效果直授沿 _weaken_accumulate_probe 前例。 func _hero_attack_row_with(opts: Dictionary, hero_atk_val: int, seed_val: int) -> Dictionary: var rng = load("res://core/rng_service.gd").new() rng.set_master_seed(seed_val) diff --git a/game/tests/unit/battle_settlement/round_flow_skill_pool_test.gd b/game/tests/unit/battle_settlement/round_flow_skill_pool_test.gd index 677a9f25..2542adac 100644 --- a/game/tests/unit/battle_settlement/round_flow_skill_pool_test.gd +++ b/game/tests/unit/battle_settlement/round_flow_skill_pool_test.gd @@ -464,7 +464,8 @@ func _row_phase(row: Dictionary, ri: int) -> int: "damage": return 2 # 案C：全部伤害兑付=战斗段（主角先手/敌方还手/地图账殿后统一结账） "reward_ineffective": - return 2 if String(row.get("forfeit", "")) == "hero_down" else 0 # 战斗段末地图账不兑留痕 vs 授予段未生效 + # 战斗段末地图账行（hero_down 兜底留痕 / deferred 直伤无目标——行载荷自描述键，迭代 48 卡①裁决补记） + return 2 if String(row.get("forfeit", "")) == "hero_down" or bool(row.get("deferred", false)) else 0 # 战斗段末地图账不兑留痕 vs 授予段未生效 match String(row.get("source_class", "")): "safe_cell", "mine_hit", "seal_hit", "seal_miss", "clear_reward": if int(row.get("round_index", ri)) < ri: diff --git a/game/tests/unit/data_config/table_structure_guard_test.gd b/game/tests/unit/data_config/table_structure_guard_test.gd index 7a06fa7a..ff0d7f8b 100644 --- a/game/tests/unit/data_config/table_structure_guard_test.gd +++ b/game/tests/unit/data_config/table_structure_guard_test.gd @@ -269,10 +269,10 @@ func _smoke_ac_swap_guard() -> void: # ① 三处无双轨 Kit.check("状态表 defense_up 行已退役（无双轨）", _db.combat.status_effects.row_by_id("defense_up").is_empty()) var ac_row: Dictionary = _db.combat.status_effects.row_by_id("ac_up") - Kit.check("ac_up 行在位（hero · +1 ac · 只续时不叠层——2026-09-20 主角增益续时批）", + Kit.check("ac_up 行在位（hero · +1 ac · 累加语义——同名重复授予加时长不叠层）〔状态累加 V4〕", not ac_row.is_empty() and String(ac_row.get("target_side", "")) == "hero" \ and String(ac_row.get("unit", "")) == "ac" and int(ac_row.get("magnitude", 0)) == 1 \ - and String(ac_row.get("stack_policy", "")) == "refresh") + and String(ac_row.get("stack_policy", "")) == "accumulate") var ac_reward := {} for row in _db.run.reward.rows: if String(row.get("cell_type", "")) == "defense" and String(row.get("polarity", "")) == "positive": diff --git a/reports/需求卡片-状态合并累加制批-2026-11-01.md b/reports/需求卡片-状态合并累加制批-2026-11-01.md index d0e59944..c13fa861 100644 --- a/reports/需求卡片-状态合并累加制批-2026-11-01.md +++ b/reports/需求卡片-状态合并累加制批-2026-11-01.md @@ -2,6 +2,7 @@ > 日期：2026-11-01 > 性质：需求改动实现批卡片；零待裁点，开工门＝U0。 +> 裁决补记（2026-11-02 用户甲案）：round_flow 相位单调断言红＝测试分类器对「战斗段地图账未生效行」缺映射（生产行序符合案C）——补记第十一节申报第 5 条＋第七节写面两行＋第十节 T1 两步；方案＝生产行载荷自描述键 `deferred`（零行为改动、观测面扩展），裁决原文见 reports/iter48-停机呈报-卡1状态相位断言-2026-11-02.md。 > 来源：待办清单「状态合并改累加制——同名状态重复授予改加时长」条（2026-10-31 用户立案并四点裁决闭合；登记批 git 提交 d216ae04，设计闭合批 870b2f2b）。 > 唯一事实源：[增益状态](../design/gdd/systems/战斗循环/增益状态.md) v1.3（2026-10-31）——§同名状态再拿到（合并策略表：叠加／累加／并存）· §剩余次数怎么倒数 · 「幅度与量纲」节盾牌加护行与字段表。本卡只实现上述设计，不改设计。 > 估时：T0–T2 ≈ **1.0～1.25 个工作日**（中值 ≈1.125）。 @@ -79,6 +80,8 @@ | game/tests/unit/battle_settlement/roll_multiplier_test.gd | 掷点/削弱/AC11 套件 | 行为改动（测试面）：削弱累加探针换代＋新增三条断言（对照表 R2–R4；申报见第十一节） | | game/tests/unit/data_config/table_structure_guard_test.gd | 表结构守卫 | 行为改动（测试面）：ac_up 行断言换代（对照表 R1；申报见第十一节） | | game/tests/README.md | 能力清单与覆盖域表 | 写卡段维护件——本卡落位时已入表，实现批零改动 | +| game/core/damage_controller.gd `_ineffective_row` | 行载荷加自描述键 `deferred: true`（战斗段地图账未生效行——裁决补记 2026-11-02 甲案） | 非行为改动（观测面扩展，零绘制/结算变化；对照表 R7） | +| game/tests/unit/battle_settlement/round_flow_skill_pool_test.gd `_row_phase` | reward_ineffective 分类补 deferred 分支（战斗段）——裁决补记随动（对照表 R7） | 行为改动（测试面，断言判据随申报 5 换代） | ## 八、落位与可复用性预检 @@ -107,6 +110,7 @@ - [ ] roll_multiplier_test.gd：③段 :157–158 注释与 :160–161 断言换代（重置 2 → 累加 3）；`_weaken_refresh_probe` 改名 `_weaken_accumulate_probe`（:159 调用点与 :342 引用注释随改），探针改五段式——授予(2)→衰减(1)→重授(1+2=3，V1)→覆盖值重授(3+3=6，V3)→同轮再授(6+2=8，V2)；返回 {layers, pool, after_regrant, after_override, after_extra, snap_magnitude} 六读数（pool 键保留——:163 池读数断言零改动在位） - [ ] 新增断言三条（尾缀〔状态累加 V1/V2/V3〕）：V1＝重授后 remaining==3、layers==1、snap_magnitude==−65（首授快照恒定）；V3＝覆盖值累加后 remaining==6；V2＝同轮再授后 remaining==8 - [ ] table_structure_guard_test.gd:273–275：ac_up 行断言 `stack_policy == "accumulate"`＋消息换代（累加语义句）——可核查：断言尾缀〔状态累加 V4〕可 grep、套件绿 +- [ ] 裁决补记随动（2026-11-02 甲案）：`_ineffective_row` 载荷加 `"deferred": true`；`_row_phase` reward_ineffective 分类补 `or bool(row.get("deferred", false))` → 战斗段 - [ ] 孤立验证：六套件最窄复跑（roll_multiplier／table_structure_guard／system_contracts／damage_controller／seal_polarity／round_flow_skill_pool——EV 锚套件提前曝红）全绿或按 L3②停机、断言尾缀可 grep **T2 产出物落位（0.25 天）** @@ -126,6 +130,7 @@ | R4 | 覆盖时长按覆盖值累加 | apply_effect :37 override 入累加臂 | 〔状态累加 V3〕（覆盖值 3 重授 remaining==6） | 单元 | 同上（真实施放链 combat_segment :89 走同口） | remaining 累加覆盖值 | 新增：同名状态合并策略（叠加/累加/并存）与剩余次数语义 | | R5 | 快照纪律不变（首授快照恒定） | 非行为改动——累加臂沿用既有重授不换快照口径，零代码变化 | 〔状态累加 V1〕快照断言（snap_magnitude==−65） | 单元 | 同 R2 | 快照域取值（snap_magnitude） | 新增：同名状态合并策略（叠加/累加/并存）与剩余次数语义 | | R6 | 叠加/并存两臂零变化 | 非行为改动——两臂零代码变化 | damage_controller_test :391＋seal_polarity_test :684＋system_contracts_test 布尔受理——既有断言零改动复跑即回归证词 | 单元 | 各既有真实入口 | 既有断言零改动 | 既有：系统间契约词表与受理口径：五配方互异／掷点标签成员性／增益布尔受理 | +| R7 | 战斗段地图账未生效行归类随动（裁决补记 2026-11-02 甲案） | damage_controller.gd `_ineffective_row` 载荷加 `deferred: true`（非行为改动）；round_flow `_row_phase` 分类补 deferred→战斗段 | round_flow_skill_pool_test 相位单调断言（同轮执行且通过——seed 42 轨迹含该行型）＋五套件回归 | 单元 | `_run_battle(1,0)` 真实结算流 | 相位分类器按自描述键归类；既有申报 1–4 与 R1–R6 断言零变化 | 既有：战斗屏三区几何与棋盘格边长（该套件承载面）——行载荷观测面扩展不立新能力行 | **测试削弱申报**（允许的旧断言改动，全部为语义换代非削弱）： @@ -133,6 +138,7 @@ 2. roll_multiplier_test.gd:157–158 注释、:230 探针注释与 :342 探针名引用注释换代（「续时：remaining ← 2」→累加口径；旧函数名 `_weaken_refresh_probe` 全仓清零）。 3. table_structure_guard_test.gd:274–275 ac_up 行断言值与消息换代（refresh→accumulate）——同上。 4. damage_controller_test.gd:376 断言标签「refresh 政策重复授予受理且层数恒 1（叠层改续时裁决）——池值不变」——行为面累加制下仍绿（层数恒 1、池值不变，断言数值零改动），仅标签文字换代为累加口径。 +5. **round_flow_skill_pool_test.gd `_row_phase` 分类补 deferred 分支（裁决补记 2026-11-02 用户甲案）**：「每轮行相位单调」断言的相位分类器对「战斗段地图账未生效行」（damage_controller `_ineffective_row` 发射、案C 口径在案）原判第 0 段——累加制改变战斗轨迹后该行型首次在种子 42 轨迹出现致断言红。随动＝`_ineffective_row` 行载荷加 `deferred: true` 自描述键（非行为改动）＋分类器按键归战斗段（相位 2）。五段全序不变量强度不降：归类的是既有设计口径行型，非放宽任何阈值。裁决原文＝reports/iter48-停机呈报-卡1状态相位断言-2026-11-02.md。 ## 十二、异常分级