[Barbatosishere]

修复内容

1. 消消乐 T/L 型漏消（Match3GameScreen + 新增 Match3Matches）

findAllMatches 抽成无 Minecraft 依赖的纯函数 Match3Matches.findAll：原实现用 checked[][] 去重，T/L 交叉型的竖臂/短边因交叉点已计入而被整段漏消（玩家看不到的幽灵方块残留在棋盘上）。新增 grid[x][y] >= 0 守卫。
回归测试：match3FindsVerticalArmOfTShape（T 横杠+竖臂、L 横杠+短边、旁邻格不误消）。

2. 推箱子零箱局与生成鲁棒性（SokobanScreen）

• 生成重试 10→24 次；仍失败时 forcePlayableLevel 写入一箱一目标的保底关，杜绝零箱局落到玩家手里
• 玩家落点只落在空地，不再覆盖箱子/目标（原来会凭空少箱甚至开局即通）
• 打乱中擦到墙角的推动改为跳过该次推动，不再整关作废（小地图上整关作废几乎必然耗尽重试次数）
• 被推上目标点的 '+' 保留目标语义；收尾从未推动的原始 '+' 箱+目标一起撤
• 终局按盘面重数箱/目标数并校验一致，不通关的局直接重生成

3. 俄罗斯方块连消中断（TetrisGameScreen）

消行音效调用移到 y--（连续满行重检测）之后，并缓存 Minecraft.getInstance()——player 为空时不再中断连续消行逻辑。

验证

• MiniGameRegressionTest 4/0 通过（含新增 Match3 用例）
• 无头压测 harness（.bench/，未随库提交）全量复跑：MiniGamesStress 26270/0 等全部 PASS

[kilo-code-bot]

Code Review Summary

This review did not finish. The model reached its output limit before it
could write the review — a reasoning model can spend the whole budget thinking.
Re-run the review, or lower the model's thinking effort, and it should get
further. Any inline comments below are from an earlier review.

Previous Review Summaries (5 snapshots)

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review

This review did not finish. The model reached its output limit before it
could write the review — a reasoning model can spend the whole budget thinking.
Re-run the review, or lower the model's thinking effort, and it should get
further. Any inline comments below are from an earlier review.

Previous review

This review did not finish. The model reached its output limit before it
could write the review — a reasoning model can spend the whole budget thinking.
Re-run the review, or lower the model's thinking effort, and it should get
further. Any inline comments below are from an earlier review.

Previous review

This review did not finish. The model reached its output limit before it
could write the review — a reasoning model can spend the whole budget thinking.
Re-run the review, or lower the model's thinking effort, and it should get
further. Any inline comments below are from an earlier review.

Previous review

This review did not finish. The model reached its output limit before it
could write the review — a reasoning model can spend the whole budget thinking.
Re-run the review, or lower the model's thinking effort, and it should get
further. Any inline comments below are from an earlier review.

Previous review

This review did not finish. The model reached its output limit before it
could write the review — a reasoning model can spend the whole budget thinking.
Re-run the review, or lower the model's thinking effort, and it should get
further. Any inline comments below are from an earlier review.

---

_{Reviewed by free · Input: 0 · Output: 0 · Cached: 0}