zizi c99014c243 test(aigc-3b-B): 派发面 e2e 真证安全面+全自动管线闭合(HMAC双向),P5 生成质量转 L1
3b-B e2e(mini-desktop 真跑·证据 wg1/gen-worker/evidence-3bB/):

 P0-P4 安全面+全自动管线=真证闭合(铁证 backend-callback-hmac.log + p4*.txt):
- HMAC 双向证:缺签名头→HTTP 401 / 错签(deadbeef≠482147b5)→HTTP 401 / 正签→过门(走到下游"任务不存在"非401)
  =关死 /dify/callback-internal 裸 @PermitAll 缺口
- 全自动派发链真通:执行器认领 taskId→dispatchGeneric 投 §6.1 job→worker 真生成(真烧模型)→
  HMAC 签名回调→验签通过→handleCallback 落库(gen-9303/9304/9305 三轮齐全)

 P5 一句话真生成=未达终局:三款全 nine_gate_failed。但截图(gen-9303-after-play.png)证生成
  真出 on-theme 可渲可动的「点击小怪物」游戏(Score/3s 倒计时/多彩怪物 monsters/targetScore=12),
  仅点击→计分坏(Score 仍 0)→九门确定性地板正确拦截「看着像但玩不了」=质量门正确履职、拒发不可玩。

判:L0 交付(管线+安全+质量门)已完成且证;缺=便宜模型把「点击命中→计分」写对=L1 生成域
  (prompt 点击坑+模型档+player/calibration,run_studio 自修 max_repairs 未救回)→转 L1 v3 迭代。

附带:serve-and-play.sh 跨平台 Chrome 探测(CHROME_BIN 覆盖>Linux /usr/bin>macOS>PATH 兜底,
  Linux 在前因 mini-desktop x86 是权威 e2e 机);p3bB_*.sql 发布四态/翻 feed 桩。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-14 16:20:28 +00:00

133 lines
4.9 KiB
JSON
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

{
"game_id": "gen-9303",
"stage": "stage1",
"pass": false,
"repairs": 5,
"player_rounds_used": 0,
"design_text": "你是轻量 H5 小游戏的【设计 agent】。输入是“限时点击不断冒出的小怪物得分”输出设计稿。\n\n---\n\n## 设计稿\n\n1. **核心机制**:每 1.2~1.8 秒从屏幕顶部或两侧随机刷出一只小怪物(大小约 80×80怪物缓慢下落或横向飘移。玩家需在 30 秒内点击怪物得分,点击空白处或不点击不会惩罚,但得分过少则输。\n2. **胜负条件**\n - **赢**:倒计时结束时,得分 ≥ 目标分 12。\n - **输**:倒计时结束得分 < 12或点击到随机出现的炸弹怪物每 8~10 秒混入一只)立即输。\n3. **操作**:触摸点击怪物身体区域(检测点击位置与怪物矩形中心 ≤ 40px。无连续操控。\n4. **关键数值**\n - 怪物出现间隔1.2~1.8 秒(随机),首个怪物在 1.0 秒后出现。\n - 怪物移动速度:垂直下落 40~60 px/s或水平飘移 30~50 px/s方向随机。\n - 炸弹怪物出现概率:每 8~10 秒一次,点击即游戏结束。\n - 倒计时30 秒,显示在右上角。\n - 目标分12 分。\n5. **Juice 反馈**\n - 点击普通怪物 → 粒子爆发8~12 个彩色碎片向外扩散 20px+ 轻快叮声(必要)。\n - 点击炸弹 → 全屏红色闪烁 + 短时震动200ms+ 低音爆炸音效。\n - 倒计时最后 5 秒:加速跳动提示。\n6. **画面要点**\n - 背景:浅绿色渐变(#D4E9D5 → #A8D8A8全屏铺满。\n - 怪物:圆形/椭圆形,随机颜色(橙/紫/蓝/粉),眼睛是两个小黑点,简单可爱。\n - 炸弹:黑白圆形,直径 70带引线图案。\n - HUD左上角分数白色粗体 36px右上角倒计时红色 32px最后 5 秒变大)。\n7. **可观测进展**\n - 分数实时增加(点击屏幕左上角数字上升)。\n - 倒计时数字递减,最后 5 秒闪烁。\n - 怪物被点击后消失(粒子动画证明击中)。\n - 炸弹出现时屏幕边缘出现红色闪烁提示。\n - 胜负切换:倒计时归零 → 切换至结算界面(显示得分+胜/败字样),点击任意处可重玩。\n\n```gatespec\n{\n \"exportState\": [\"phase\", \"score\", \"timeLeft\", \"monsters\"],\n \"driver\": {\"type\": \"none\"},\n \"controlCheck\": null,\n \"assertAfterPlay\": [\n {\"path\": \"score\", \"op\": \">=\", \"why\": \"点击怪物后得分增加\"}\n ],\n \"expectLatch\": true\n}\n```",
"gatespec": {
"exportState": [
"phase",
"score",
"timeLeft",
"monsters"
],
"driver": {
"type": "none"
},
"controlCheck": null,
"assertAfterPlay": [
{
"path": "score",
"op": ">=",
"why": "点击怪物后得分增加"
}
],
"expectLatch": true
},
"attempts": [
{
"attempt": 0,
"role": "code",
"model": "deepseek-v4-flash",
"stage_fail": "seven_gate",
"usage": {
"in": 4206,
"out": 5531
},
"seven_gate_pass": false,
"guards": {}
},
{
"attempt": 1,
"role": "fix",
"model": "deepseek-v4-flash",
"stage_fail": "seven_gate",
"usage": {
"in": 4249,
"out": 6241
},
"seven_gate_pass": false,
"guards": {}
},
{
"attempt": 2,
"role": "fix",
"model": "deepseek-v4-flash",
"stage_fail": "seven_gate",
"usage": {
"in": 4249,
"out": 5585
},
"seven_gate_pass": false,
"guards": {}
},
{
"attempt": 3,
"role": "fix",
"model": "deepseek-v4-flash",
"stage_fail": "seven_gate",
"usage": {
"in": 4249,
"out": 5575
},
"seven_gate_pass": false,
"guards": {}
},
{
"attempt": 4,
"role": "fix",
"model": "deepseek-v4-flash",
"stage_fail": "seven_gate",
"usage": {
"in": 4249,
"out": 6523
},
"seven_gate_pass": false,
"guards": {}
},
{
"attempt": 5,
"role": "fix",
"model": "deepseek-v4-flash",
"stage_fail": "seven_gate",
"usage": {
"in": 4249,
"out": 5490
},
"seven_gate_pass": false,
"guards": {}
}
],
"seven_gate_verdict": null,
"player": null,
"tokens": {
"prompt": 26160,
"completion": 36307
},
"tokens_by_model": {
"deepseek-v4-flash": [
26160,
36307
]
},
"cost": {
"by_model": {
"deepseek-v4-flash": {
"in": 26160,
"out": 36307,
"rmb": 0.07857
}
},
"total_rmb": 0.07857,
"gate_rmb": 0.15
},
"wall_s": 388.4,
"models": {
"design": "deepseek-v4-flash",
"code": "deepseek-v4-flash",
"fix": "deepseek-v4-flash",
"player_vision": "MiniMax-M3",
"player_text": "deepseek-v4-flash"
}
}