初始提交：3 款纯音频游戏 + 共享引擎 + 完整文档与资产管线

- 共享引擎（app/src/main）：GestureEngine 6 手势识别、AudioEngine 三层音频
  （ExoPlayer BGM 交叉淡入 / 对白 ducking / SoundPool 低延迟）、Game/GameEngine
  状态机接口、MainActivity 全屏封面 + TalkBack
- 《声命线》storyline：末世互动广播剧，3 幕 8 抉择点 4 结局
- 《声东击西》reflex：听声辨位反应，12 波战斗 + BGM 三档 BPM 变速
- 《暗室》escape：3x3 网格密室逃脱，钥匙→铁柜→保险丝→电箱→大门谜题链
- 文档：setup（从零搭建）/ architecture（架构）/ engine-api（引擎 API）/
  asset-pipeline（资产流水线）/ games/*（三款游戏设计文档）
- 资产管线：Qwen3-TTS 对白 + 程序合成音效/BGM + Mureka BGM 的 gen 脚本与台账，
  全部 ogg 资产入库可复现
- env.d 环境脚本：android/tts 入库；sfx/mureka 含密钥不入库，仅提交 .example.sh 模板

tools/stable-audio-3/tests/utils/audio.py

@@ -0,0 +1,50 @@
+import math
+
+import torch
+
+AMPLITUDE = 0.8
+
+
+def sine_wave(
+    duration: float,
+    sample_rate: int,
+    channels: int = 2,
+    freq: float = 440.0,
+    device: str = "cpu",
+    half: bool = False,
+) -> torch.Tensor:
+    """Return a [channels, samples] sine-wave tensor in [-AMPLITUDE, AMPLITUDE]."""
+    t = torch.linspace(0, duration, int(duration * sample_rate), device=device)
+    wave = AMPLITUDE * torch.sin(2 * math.pi * freq * t)
+    if half:
+        wave = wave.half()
+    return wave.expand(channels, -1).clone()
+
+
+def assert_audio_valid(
+    audio: torch.Tensor, expected_duration: float, sample_rate: int
+) -> None:
+    """Assert that an audio tensor looks like valid, non-trivial audio output.
+
+    Args:
+        audio: Output tensor of shape [B, C, T].
+        expected_duration: Expected length in seconds.
+        sample_rate: Sample rate in Hz.
+    """
+    assert audio is not None, "Audio output is None"
+    assert audio.ndim == 3, f"Expected 3D tensor [B, C, T], got shape {audio.shape}"
+
+    expected_samples = int(expected_duration * sample_rate)
+    actual_samples = audio.shape[-1]
+    tolerance = 0.15
+    relative_error = abs(actual_samples - expected_samples) / expected_samples
+    assert relative_error < tolerance, (
+        f"Audio length {actual_samples} samples deviates more than {tolerance * 100:.0f}% "
+        f"from expected {expected_samples} samples ({expected_duration}s @ {sample_rate}Hz)"
+    )
+
+    assert not torch.isnan(audio).any(), "Audio output contains NaN values"
+
+    abs_max = audio.abs().max().item()
+    assert abs_max > 0.001, f"Audio output appears silent (abs_max={abs_max:.5f})"
+    assert abs_max <= 1.0, f"Audio output is clipping (abs_max={abs_max:.5f})"