《衍》是一款单机、确定性、网格制的战棋／解谜原型。棋子按自身规则自动行动，玩家只能替换一枚棋子当前一动的行动意图；目标是先读懂无人干涉时的因果链，再用很少的干涉改变后续演化。
Yan is a deterministic single-player grid tactics and puzzle prototype. Pieces act automatically under shared rules. The player may replace one piece's current intent, first reading the uninterrupted causal chain and then changing its evolution with minimal intervention.
个人工作：定义白板棋子规则、玩家干涉边界、固定水晶与替代性战斗；在 Unity 中完成确定性模拟、关卡数据、关卡编辑器、无文字交互界面、撤回／重开和两步教学流程。
Role: Defined the white-piece rules, player intervention boundary, crystal goal, and replacement combat event; implemented the deterministic simulation, level data, editor, wordless interface, undo/restart, and two-step tutorial in Unity.
阶段状态：2026-08-10 完成两步教学垂直切片；不是完整关卡集。
Stage status: two-step tutorial vertical slice completed on 2026-08-10; not a complete level set.
DESIGN DECISIONS / TEACHING ORDER
- 替换意图玩家只能从棋子拥有的前进、左转、右转和掉头中选择，不直接移动棋子或指定战斗结果。
- 替代性战斗敌对占位重叠时，本动原意图被一次同时伤害交换替代；下一动重新计算，不保存隐藏的持续交战状态。
- 先教目标教学关 0 无需干涉，先展示自动推进、三次命中与摧毁水晶；教学关 1 再引入一次右转。
- 缩短收尾教学关 1 从 V0.1 调整为 V0.2，减少解开后的机械前进，保留“先观察，再找到破局点”的节拍。
- REPLACE INTENTThe player chooses only among the piece's own forward, left, right, and turn-around intents, never a direct result.
- REPLACEMENT COMBATOpposing occupancy replaces the current intent with one simultaneous damage exchange; the next step is recalculated without a hidden combat state.
- TEACH GOAL FIRSTTutorial 0 needs no intervention and demonstrates automatic progression, three hits, and crystal destruction before Tutorial 1 adds one right turn.
- SHORTEN THE TAILTutorial 1 V0.2 removes mechanical walking after the solution while preserving the observe-then-intervene beat.
