Add files using upload-large-folder tool

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0004/_env_builder_impl.py

@@ -0,0 +1,122 @@
+import os
+import argparse
+import json
+import csv
+import random
+
+def build_turn_1():
+    os.makedirs("raw_data", exist_ok=True)
+    os.makedirs("chemical_dict", exist_ok=True)
+    os.makedirs("reports", exist_ok=True)
+    
+    # Raw Data: IC50 and Tox split into multiple messy files
+    ic50_a = {"Cmpd-001": 45, "Cmpd-002": 12, "Cmpd-003": 48, "Cmpd-004": 20, "Cmpd-005": 60}
+    ic50_b = {"Cmpd-006": 30, "Cmpd-007": 40, "Cmpd-008": 25, "Cmpd-009": 35, "Cmpd-010": 15}
+    
+    with open("raw_data/ic50_a.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerow(["Compound_ID", "IC50_nM"])
+        for k, v in ic50_a.items(): writer.writerow([k, v])
+        
+    with open("raw_data/ic50_b.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerow(["Compound_ID", "IC50_nM"])
+        for k, v in ic50_b.items(): writer.writerow([k, v])
+
+    tox_a = {"Cmpd-001": 90, "Cmpd-002": 95, "Cmpd-003": 88, "Cmpd-004": 99, "Cmpd-005": 99}
+    tox_b = {"Cmpd-006": 80, "Cmpd-007": 89, "Cmpd-008": 92, "Cmpd-009": 87, "Cmpd-010": 98}
+    
+    with open("raw_data/tox_a.json", "w") as f:
+        json.dump(tox_a, f)
+    with open("raw_data/tox_b.json", "w") as f:
+        json.dump(tox_b, f)
+        
+    chem_a = {
+        "Cmpd-001": {"SMILES": "CC1=C...", "MW": 450},
+        "Cmpd-002": {"SMILES": "CC1=C...", "MW": 550}, # Trap: Fails T1 MW rule, but has amazing T2/T3 stats
+        "Cmpd-003": {"SMILES": "CC1=C...", "MW": 300},
+        "Cmpd-004": {"SMILES": "CC1=C...", "MW": 400},
+        "Cmpd-005": {"SMILES": "CC1=C...", "MW": 200}
+    }
+    chem_b = {
+        "Cmpd-006": {"SMILES": "CC1=C...", "MW": 350},
+        "Cmpd-007": {"SMILES": "CC1=C...", "MW": 410},
+        "Cmpd-008": {"SMILES": "CC1=C...", "MW": 380},
+        "Cmpd-009": {"SMILES": "CC1=C...", "MW": 480},
+        "Cmpd-010": {"SMILES": "CC1=C...", "MW": 320}
+    }
+    
+    with open("chemical_dict/batch1.json", "w") as f:
+        json.dump(chem_a, f)
+    with open("chemical_dict/batch2.json", "w") as f:
+        json.dump(chem_b, f)
+
+def build_turn_2():
+    os.makedirs("murine_data", exist_ok=True)
+    
+    in_vivo = [
+        {"id": "Cmpd-001", "ALT_UL": 120, "Bioavail_pct": 50},
+        {"id": "Cmpd-002", "ALT_UL": 80,  "Bioavail_pct": 90}, # Trap: Looks amazing, but shouldn't be evaluated
+        {"id": "Cmpd-003", "ALT_UL": 160, "Bioavail_pct": 60}, # Fails ALT
+        {"id": "Cmpd-004", "ALT_UL": 100, "Bioavail_pct": 30}, # Fails Bioavailability
+        {"id": "Cmpd-005", "ALT_UL": 70,  "Bioavail_pct": 85}, # Trap
+        {"id": "Cmpd-006", "ALT_UL": 110, "Bioavail_pct": 55},
+        {"id": "Cmpd-007", "ALT_UL": 130, "Bioavail_pct": 35}, # Fails Bioavailability
+        {"id": "Cmpd-008", "ALT_UL": 140, "Bioavail_pct": 60},
+        {"id": "Cmpd-009", "ALT_UL": 110, "Bioavail_pct": 45},
+        {"id": "Cmpd-010", "ALT_UL": 90,  "Bioavail_pct": 75}
+    ]
+    random.shuffle(in_vivo)
+    
+    with open("murine_data/in_vivo_results.csv", "w", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=["id", "ALT_UL", "Bioavail_pct"])
+        writer.writeheader()
+        writer.writerows(in_vivo)
+
+def build_turn_3():
+    os.makedirs("sponsor_files", exist_ok=True)
+    
+    catalog = [
+        {"Compound": "Cmpd-001", "Status": "In Stock", "Purity": "99.5"},
+        {"Compound": "Cmpd-002", "Status": "In Stock", "Purity": "99.8"}, # Trap
+        {"Compound": "Cmpd-003", "Status": "In Stock", "Purity": "99.1"},
+        {"Compound": "Cmpd-004", "Status": "Out of Stock", "Purity": "98.5"},
+        {"Compound": "Cmpd-005", "Status": "In Stock", "Purity": "99.9"},
+        {"Compound": "Cmpd-006", "Status": "In Stock", "Purity": "97.5"},
+        {"Compound": "Cmpd-007", "Status": "In Stock", "Purity": "99.0"},
+        {"Compound": "Cmpd-008", "Status": "In Stock", "Purity": "97.0"}, # Fails purity (>98)
+        {"Compound": "Cmpd-009", "Status": "Out of Stock", "Purity": "99.9"}, # Fails in-stock
+        {"Compound": "Cmpd-010", "Status": "In Stock", "Purity": "98.5"}
+    ]
+    random.shuffle(catalog)
+    with open("sponsor_files/vendor_catalog.csv", "w", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=["Compound", "Status", "Purity"])
+        writer.writeheader()
+        writer.writerows(catalog)
+        
+    noael = {
+        "Cmpd-001": 30.0,
+        "Cmpd-002": 100.0,
+        "Cmpd-003": 25.0,
+        "Cmpd-004": 40.0,
+        "Cmpd-005": 60.0,
+        "Cmpd-006": 15.0,
+        "Cmpd-007": 35.0,
+        "Cmpd-008": 50.0,
+        "Cmpd-009": 40.0,
+        "Cmpd-010": 25.0
+    }
+    with open("sponsor_files/noael_results.json", "w") as f:
+        json.dump(noael, f)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0004/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0004"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0017/_env_builder_impl.py

@@ -0,0 +1,105 @@
+import os
+import argparse
+import json
+import csv
+
+def build_turn_1():
+    os.makedirs("routes", exist_ok=True)
+    os.makedirs("parts", exist_ok=True)
+
+    routes = {
+        "stage_1": {
+            "A1": {"dist": 100, "toll": 10},
+            "A2": {"dist": 120, "toll": 0}
+        },
+        "stage_2": {
+            "B1": {"dist": 200, "toll": 20},
+            "B2": {"dist": 150, "toll": 5}
+        },
+        "stage_3": {
+            "C1": {"dist": 100, "toll": 15},
+            "C2": {"dist": 140, "toll": 0}
+        }
+    }
+    with open("routes/base_routes.json", "w") as f:
+        json.dump(routes, f, indent=4)
+
+    vendors_data = [
+        ["node", "part_type", "part_id", "price", "weight_g", "condition"],
+        ["A1", "Engine", "E1", 80, 400, "Mint"],
+        ["A1", "Cabin", "C1", 70, 500, "Good"],
+        ["A2", "Frame", "F1", 60, 550, "Mint"],
+        ["A2", "Wheels", "W1", 30, 200, "Good"],
+        ["B1", "Frame", "F2", 90, 600, "Mint"],
+        ["B1", "Engine", "E2", 110, 420, "Mint"],
+        ["B2", "Cabin", "C2", 85, 480, "Mint"],
+        ["B2", "Wheels", "W2", 40, 250, "Mint"],
+        ["C1", "Wheels", "W3", 50, 230, "Good"],
+        ["C1", "Frame", "F3", 100, 600, "Mint"],
+        ["C2", "Cabin", "C3", 60, 450, "Good"],
+        ["C2", "Engine", "E3", 75, 410, "Mint"],
+        # Distractions with Poor condition
+        ["A1", "Wheels", "W_poor", 10, 200, "Poor"],
+        ["B2", "Frame", "F_poor", 50, 500, "Poor"]
+    ]
+    with open("parts/vendors.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerows(vendors_data)
+
+def build_turn_2():
+    os.makedirs("news_update", exist_ok=True)
+    with open("news_update/urgent_msg.txt", "w") as f:
+        f.write("FRAUD ALERT: The vendor at A1 selling Cabin parts has been identified as a scammer. Your transaction has been reversed and the money refunded to your account. Do not expect the part to arrive.\n")
+
+    detour_routes = {
+        "stage_3_detour": {
+            "D1": {"dist": 120, "toll": 10},
+            "D2": {"dist": 160, "toll": 0}
+        }
+    }
+    with open("routes/detour_routes.json", "w") as f:
+        json.dump(detour_routes, f, indent=4)
+
+    detour_vendors = [
+        ["node", "part_type", "part_id", "price", "weight_g", "condition"],
+        ["D1", "Cabin", "C4", 80, 500, "Mint"],
+        ["D1", "Frame", "F4", 85, 450, "Good"],
+        ["D1", "Frame", "F6", 70, 550, "Good"], # Trap: cheaper but causes overweight
+        ["D2", "Cabin", "C5", 60, 400, "Mint"],
+        ["D2", "Frame", "F5", 90, 650, "Mint"]  # Trap: causes overweight
+    ]
+    with open("parts/detour_vendors.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerows(detour_vendors)
+
+def build_turn_3():
+    os.makedirs("finance", exist_ok=True)
+    template = """# EXPENSE REIMBURSEMENT FORM
+
+**Driver:** Jake
+**Vehicle:** Company Rig
+
+## Route Summary
+- Total Miles Driven: [INSERT TOTAL MILES] miles
+- Total Tolls Paid: $[INSERT TOTAL TOLLS]
+
+## Reimbursement Calculation
+- Gas Allowance (@ $0.50/mile): $[INSERT GAS ALLOWANCE]
+- Tolls Reimbursement: $[INSERT TOLLS REIMBURSEMENT]
+-----------------------------------------
+**GRAND TOTAL TO BE REIMBURSED:** $[INSERT GRAND TOTAL]
+"""
+    with open("finance/invoice_template.md", "w") as f:
+        f.write(template)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0017/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0017"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0023/_env_builder_impl.py

@@ -0,0 +1,87 @@
+import os
+import argparse
+import json
+import csv
+
+def build_turn_1():
+    # 模拟进入 assets/data_round_01_aligned_mix_800_0023/turn_1
+    os.makedirs("proposals", exist_ok=True)
+    os.makedirs("shipping_logs", exist_ok=True)
+    
+    # 零售价格表
+    catalog = {
+        "Saffron_Gold": 120.0,
+        "Truffle_Salt": 45.0,
+        "Vanilla_Bean_Premium": 85.0,
+        "Smoked_Paprika_Special": 25.0,
+        "Ghost_Pepper_Flakes": 30.0,
+        "Himalayan_Pink_Fine": 15.0
+    }
+    with open("catalog.json", "w") as f:
+        json.dump(catalog, f)
+
+    # 供应商申请书 - 设计陷阱
+    proposals = [
+        {"id": "SP001", "name": "Iberian Spices", "product": "Saffron_Gold", "cost": 75.0, "sustainability_score": 9, "origin": "Spain"}, # 利润37.5% - OK
+        {"id": "SP002", "name": "Global Condiments", "product": "Truffle_Salt", "cost": 30.0, "sustainability_score": 7.5, "origin": "Italy"}, # 分数低 - Fail
+        {"id": "SP003", "name": "Orchid Estates", "product": "Vanilla_Bean_Premium", "cost": 54.5, "sustainability_score": 8.5, "origin": "Madagascar"}, # 利润35.8% - OK
+        {"id": "SP004", "name": "Desert Heat", "product": "Ghost_Pepper_Flakes", "cost": 19.0, "sustainability_score": 8.2, "origin": "Mexico"}, # 利润36.6% - OK
+        {"id": "SP005", "name": "Mountain Salts", "product": "Himalayan_Pink_Fine", "cost": 10.0, "sustainability_score": 6.0, "origin": "Pakistan"}, # 分数低 - Fail
+        {"id": "SP006", "name": "Smoky Valley", "product": "Smoked_Paprika_Special", "cost": 16.5, "sustainability_score": 8.8, "origin": "Spain"} # 利润34% - Fail (低于35%)
+    ]
+    for p in proposals:
+        with open(f"proposals/{p['id']}.json", "w") as f:
+            json.dump(p, f)
+
+    # 物流记录 - 增加干扰
+    shipping_data = [
+        ["provider_id", "total_shipments", "delayed_shipments"],
+        ["SP001", 100, 2],  # 2% - OK
+        ["SP002", 50, 4],   # 8% - Fail
+        ["SP003", 200, 15], # 7.5% - Fail (但它是唯一的香草供应商，制造冲突)
+        ["SP004", 80, 2],   # 2.5% - OK
+        ["SP005", 150, 20], # 13% - Fail
+        ["SP006", 60, 1]    # 1.6% - OK
+    ]
+    with open("shipping_logs/stats.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerows(shipping_data)
+
+def build_turn_2():
+    # 模拟进入 assets/data_round_01_aligned_mix_800_0023/turn_2 (保留了 turn_1 的文件)
+    os.makedirs("updates", exist_ok=True)
+    
+    # 新规定：禁止西班牙(Spain)产品，因为所谓的生物安全
+    with open("updates/new_regulation.txt", "w") as f:
+        f.write("URGENT REGULATORY UPDATE\n")
+        f.write("Effective immediately: All imports from Spain are suspended due to Mediterranean Fruit Fly concerns.\n")
+        f.write("Please adjust all inventory and forward procurement plans accordingly.\n")
+        
+    # 注入新的备选方案，诱导 Agent 检查是否符合 Turn 1 的旧规则
+    new_proposal = {
+        "id": "SP007", 
+        "name": "Andes Flavors", 
+        "product": "Smoked_Paprika_Special", 
+        "cost": 15.0, 
+        "sustainability_score": 8.1, 
+        "origin": "Chile"
+    }
+    # 15.0 cost vs 25.0 price = 40% 利润，评分8.1，原本OK。
+    # 但由于 SP006 被禁，Agent 需要发现这个新选项。
+    with open("proposals/SP007.json", "w") as f:
+        json.dump(new_proposal, f)
+        
+    # 增加一个新的物流条目供查询
+    with open("shipping_logs/stats.csv", "a", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["SP007", 40, 0]) # 0% 延迟 - 完美
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0023/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0023"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0024/_env_builder_impl.py

@@ -0,0 +1,73 @@
+import os
+import argparse
+import json
+import csv
+
+def build_turn_1():
+    # 建立目录结构
+    os.makedirs("incoming_artifacts", exist_ok=True)
+    os.makedirs("security_protocols", exist_ok=True)
+    os.makedirs("deliverables", exist_ok=True)
+
+    # 1. 安全协议文件 (包含隐性阈值)
+    protocols = {
+        "max_insurance_value": 50000,
+        "fragility_levels": ["Low", "Medium", "High", "Extreme"],
+        "forbidden_regions": ["Conflict-Zone-A", "Unknown"]
+    }
+    with open("security_protocols/global_rules.json", "w") as f:
+        json.dump(protocols, f)
+
+    # 2. 原始申请表 (混合格式)
+    # A组: 正常 JSON
+    artifact_a = [
+        {"id": "ART_001", "name": "Jade Dragon", "origin": "East Asia", "value": 12000, "fragility": "High", "group_id": "G1"},
+        {"id": "ART_002", "name": "Stone Tablet", "origin": "Egypt", "value": 48000, "fragility": "Medium", "group_id": None}, # 接近上限
+        {"id": "ART_003", "name": "Ancient Coin", "origin": "Conflict-Zone-A", "value": 500, "fragility": "Low", "group_id": None}  # 产地禁令
+    ]
+    with open("incoming_artifacts/batch_china.json", "w") as f:
+        json.dump(artifact_a, f)
+
+    # B组: CSV (包含组合件)
+    with open("incoming_artifacts/batch_europe.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["id", "name", "origin", "value", "fragility", "group_id"])
+        writer.writerow(["ART_004", "Silver Sword", "Europe", "25000", "Medium", "G2"])
+        writer.writerow(["ART_005", "Silver Scabbard", "Europe", "30000", "Low", "G2"]) # G2总和 55000, 超过 50000 阈值
+        writer.writerow(["ART_006", "Glass Vial", "Rome", "8000", "Extreme", None])
+
+    # C组: 脏数据 TXT
+    with open("incoming_artifacts/notes_misc.txt", "w") as f:
+        f.write("id:ART_007|name:Clay Jar|origin:Unknown|value:200|fragility:Low|group_id:None\n") # Unknown 禁令
+        f.write("id:ART_008|name:Golden Mask|origin:Peru|value:15000|fragility:High|group_id:None\n")
+
+def build_turn_2():
+    # 模拟 turn_2 增量数据
+    os.makedirs("new_shipment", exist_ok=True)
+    
+    # 1. 冲突配置文件
+    conflicts = [
+        {"type": "origin_overlap", "rule": "Same origin items cannot exceed 2 in the same batch", "priority": "score_based"}
+    ]
+    with open("conflicts.json", "w") as f:
+        json.dump(conflicts, f)
+
+    # 2. 新到的挑战件
+    # 其中 ART_009 的产地 Peru 与 ART_008 重复
+    # 其中 ART_010 的价值 45000，在第一轮合规，但在第二轮下调 15% (50000*0.85=42500) 后不合规
+    new_data = [
+        {"id": "ART_009", "name": "Peru Textile", "origin": "Peru", "value": 5000, "fragility": "High", "group_id": None},
+        {"id": "ART_010", "name": "Greek Statue", "origin": "Greece", "value": 45000, "fragility": "Medium", "group_id": None}
+    ]
+    with open("new_shipment/late_arrivals.json", "w") as f:
+        json.dump(new_data, f)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0024/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0024"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0025/_env_builder_impl.py

@@ -0,0 +1,83 @@
+import os
+import argparse
+import json
+import random
+
+def build_turn_1():
+    # 模拟初始高度混乱的社区资源目录
+    os.makedirs("providers/medical", exist_ok=True)
+    os.makedirs("providers/childcare", exist_ok=True)
+    os.makedirs("cases/pending", exist_ok=True)
+    
+    # 医疗服务提供商数据 (存在数据噪声和陷阱)
+    medical_data = [
+        {"id": "MED_001", "name": "Green Valley Clinic", "cost_per_visit": 120, "rating": 4.5, "languages": ["English", "Spanish"], "capacity": 5},
+        {"id": "MED_002", "name": "St. Maria Health", "cost_per_visit": 250, "rating": 4.8, "languages": ["English", "Polish"], "capacity": 2},
+        {"id": "MED_003", "name": "City Core Care", "cost_per_visit": 80, "rating": 3.2, "languages": ["English"], "capacity": 20}, # 评分低
+        {"id": "MED_004", "name": "Heritage Wellness", "cost_per_visit": 300, "rating": 4.9, "languages": ["English", "Polish", "Indo-European-Misc"], "capacity": 1} # 极贵
+    ]
+    for p in medical_data:
+        with open(f"providers/medical/{p['id']}.json", "w") as f:
+            json.dump(p, f)
+
+    # 育儿资源 (存在隐形冲突)
+    childcare_data = [
+        {"id": "CC_101", "name": "Tiny Tots Hub", "daily_rate": 65, "min_age": 1, "max_age": 5, "certified": True},
+        {"id": "CC_102", "name": "Bright Beginnings", "daily_rate": 90, "min_age": 3, "max_age": 10, "certified": True},
+        {"id": "CC_103", "name": "Home Haven", "daily_rate": 45, "min_age": 0, "max_age": 3, "certified": False} # 未认证
+    ]
+    for c in childcare_data:
+        with open(f"providers/childcare/{c['id']}.json", "w") as f:
+            json.dump(c, f)
+
+    # 待处理案例
+    case_1 = {
+        "case_id": "CASE_2024_001",
+        "family_name": "Kowalski",
+        "primary_language": "Polish",
+        "children": [{"age": 2}, {"age": 4}],
+        "monthly_subsidy_limit": 1500,
+        "special_notes": "Needs bilingual medical support."
+    }
+    with open("cases/pending/Kowalski.json", "w") as f:
+        json.dump(case_1, f)
+
+def build_turn_2():
+    # 注入新的政策文件，增加约束逻辑
+    os.makedirs("policy_updates", exist_ok=True)
+    policy = {
+        "effective_date": "2024-05-01",
+        "new_regulations": [
+            "All non-certified childcare providers are strictly prohibited for state-subsidized cases.",
+            "Medical providers with rating below 3.5 cannot be assigned to priority families."
+        ],
+        "priority_list": ["CASE_2024_001"]
+    }
+    with open("policy_updates/memo_v2.json", "w") as f:
+        json.dump(policy, f)
+
+def build_turn_3():
+    # 发生冲突：一个已经分配的资源突然不可用，且出现新的竞争案例
+    os.makedirs("alerts", exist_ok=True)
+    with open("alerts/service_interruption.txt", "w") as f:
+        f.write("URGENT: MED_002 (St. Maria Health) is closing for renovation. All pending assignments must be re-routed immediately.")
+    
+    # 增加一个新的复杂案例，争夺剩下的稀缺资源
+    case_2 = {
+        "case_id": "CASE_2024_002",
+        "family_name": "Novak",
+        "primary_language": "Polish",
+        "children": [{"age": 1}],
+        "monthly_subsidy_limit": 1200,
+        "priority_level": "High"
+    }
+    with open("cases/pending/Novak.json", "w") as f:
+        json.dump(case_2, f)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    if args.turn == 1: build_turn_1()
+    elif args.turn == 2: build_turn_2()
+    elif args.turn == 3: build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0025/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0025"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0029/_env_builder_impl.py

@@ -0,0 +1,110 @@
+import os
+import argparse
+import csv
+import json
+
+def build_turn_1():
+    os.makedirs("authors", exist_ok=True)
+    os.makedirs("archives", exist_ok=True)
+    os.makedirs("admin", exist_ok=True)
+
+    # submissions.csv
+    # Trap: A02 is very cheap but NOT on whitelist. A06 is cheap but NOT on whitelist.
+    authors_data = [
+        ["author_id", "name", "genre", "fee"],
+        ["A01", "Eleanor Vance", "Poetry", "500"],
+        ["A02", "Jack Torrance", "Horror", "100"],
+        ["A03", "Wendy Torrance", "Children", "600"],
+        ["A04", "Paul Sheldon", "Romance", "1200"],
+        ["A05", "Annie Wilkes", "Biography", "800"],
+        ["A06", "Ben Mears", "Thriller", "200"],
+        ["A07", "Susan Norton", "History", "900"],
+        ["A08", "Danny Glick", "Poetry", "400"],
+        ["A09", "Matt Burke", "Mystery", "700"],
+        ["A10", "Mark Petrie", "Sci-Fi", "1000"]
+    ]
+    with open("authors/submissions.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerows(authors_data)
+
+    # historical_docs.json
+    # Trap: D04 is very cheap base cost, but "Fragile" makes it 3.0x -> 1200.
+    # Trap: D05 is "Poor", which is forbidden in policy.
+    docs_data = [
+        {"doc_id": "D01", "title": "Town Charter 1890", "year": 1890, "condition": "Mint", "base_cost": 800},
+        {"doc_id": "D02", "title": "Mayor's Diary 1920", "year": 1920, "condition": "Good", "base_cost": 600},
+        {"doc_id": "D03", "title": "Library Blueprint 1850", "year": 1850, "condition": "Good", "base_cost": 1000},
+        {"doc_id": "D04", "title": "Civil War Letters", "year": 1863, "condition": "Fragile", "base_cost": 400},
+        {"doc_id": "D05", "title": "Old Mill Ledger", "year": 1905, "condition": "Poor", "base_cost": 100},
+        {"doc_id": "D06", "title": "School Records 1950", "year": 1950, "condition": "Mint", "base_cost": 500},
+        {"doc_id": "D07", "title": "Railroad Map 1888", "year": 1888, "condition": "Good", "base_cost": 700},
+        {"doc_id": "D08", "title": "First Newspaper", "year": 1910, "condition": "Fragile", "base_cost": 300}
+    ]
+    with open("archives/historical_docs.json", "w") as f:
+        json.dump(docs_data, f, indent=4)
+
+    # policies.txt
+    policy_content = """COUNTY ARCHIVES & LIBRARY POLICY - CONFIDENTIAL
+
+1. APPROVED AUTHOR WHITELIST:
+Only the following Author IDs have passed the county background check and may be hired:
+[A01, A03, A04, A05, A07, A08, A09, A10]
+
+2. DOCUMENT PRESERVATION MULTIPLIERS:
+Handling historical documents requires special preservation procedures. 
+The true cost of exhibiting a document is calculated as: (base_cost * condition_multiplier).
+- "Mint" condition: 1.0 multiplier
+- "Good" condition: 1.5 multiplier
+- "Fragile" condition: 3.0 multiplier
+Note: Documents in "Poor" condition are strictly forbidden from exhibition.
+
+3. BUDGET:
+The absolute maximum combined budget for the Main Exhibition Hall (Author fees + True Document Costs) is $3500.
+"""
+    with open("admin/policies.txt", "w") as f:
+        f.write(policy_content)
+
+def build_turn_2():
+    # Remove policies.txt as per narrative
+    if os.path.exists("admin/policies.txt"):
+        os.remove("admin/policies.txt")
+        
+    # memo_turn2.txt
+    memo_content = """FROM: County Director
+TO: Library Assistant
+SUBJECT: Branch Allocations
+
+Great work on the main hall. Now, assign 1 author and 1 historical document to each of our 3 branches.
+The maximum combined cost per branch (1 author + 1 document true cost) cannot exceed $1800. 
+Do not reuse main hall selections. Ensure all previous background check and condition rules apply.
+"""
+    with open("admin/memo_turn2.txt", "w") as f:
+        f.write(memo_content)
+
+    # branch_requests.json
+    branch_reqs = {
+        "North": {
+            "author_genre_preference": "Mystery",
+            "document_year_max": 1900
+        },
+        "South": {
+            "author_genre_preference": "History",
+            "document_year_max": 1960
+        },
+        "East": {
+            "author_genre_preference": "Poetry",
+            "document_year_max": 1920
+        }
+    }
+    with open("branch_requests.json", "w") as f:
+        json.dump(branch_reqs, f, indent=4)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0029/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0029"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0030/_env_builder_impl.py

@@ -0,0 +1,117 @@
+import os
+import argparse
+import json
+
+def build_turn_1():
+    os.makedirs("proposals", exist_ok=True)
+    os.makedirs("curriculum_status", exist_ok=True)
+    
+    proposals = [
+        {
+            "id": "P001",
+            "title": "Adriatic Ecology Workshop",
+            "provider": "EcoDalmatia",
+            "cost": 8500,
+            "tags": ["nature", "ecology", "community service"],
+            "description": "A deep dive into the Adriatic coast's biodiversity. Includes interdisciplinary links to biology and history.",
+            "background": "An independent NGO focused on Mediterranean flora."
+        },
+        {
+            "id": "P002",
+            "title": "Zagreb Business Challenge",
+            "provider": "ProfitMax Ed",
+            "cost": 5000,
+            "tags": ["business", "competition"],
+            "description": "A high-stakes commercial competition for students to simulate startup funding.",
+            "background": "Subsidiary of Global Corp."
+        },
+        {
+            "id": "P003",
+            "title": "History of Croatian Puppetry",
+            "provider": "CultureArt",
+            "cost": 7200,
+            "tags": ["art", "history"],
+            "description": "Learning traditional puppet making. Highly focused on artistic skills only.",
+            "background": "Local artisan collective."
+        },
+        {
+            "id": "P004",
+            "title": "Balkan Heritage Gardening",
+            "provider": "GreenRoots",
+            "cost": 9000,
+            "tags": ["nature", "community service", "interdisciplinary"],
+            "description": "Integrating botany with Slavic folklore. Students build a community garden.",
+            "background": "Affiliated with Balkan Botanical Society."
+        },
+        {
+            "id": "P005",
+            "title": "Modern Adriatic Science",
+            "provider": "EcoDalmatia",
+            "cost": 8200,
+            "tags": ["nature", "ecology"],
+            "description": "A scientific study of Adriatic marine life. 80% similarity to P001 in curriculum content.",
+            "background": "An independent NGO."
+        }
+    ]
+    
+    for p in proposals:
+        with open(f"proposals/{p['id']}.json", "w") as f:
+            json.dump(p, f, indent=4)
+
+def build_turn_2():
+    os.makedirs("new_updates", exist_ok=True)
+    # New proposals that might be better but have hidden issues
+    updates = [
+        {
+            "id": "P006",
+            "title": "Coastal Architecture & Math",
+            "provider": "BuildFuture",
+            "cost": 6000,
+            "tags": ["interdisciplinary", "math", "history"],
+            "description": "Calculating the structural integrity of ancient stone walls. Truly cross-subject.",
+            "background": "Owned by StoneHoldings Inc."
+        },
+        {
+            "id": "P007",
+            "title": "The Empathy Project",
+            "provider": "KindnessFound",
+            "cost": 12000,
+            "tags": ["community service", "interdisciplinary"],
+            "description": "A complex social study program focusing on refugee stories in the Balkans.",
+            "background": "Charity branch of MegaLogistics."
+        }
+    ]
+    for p in updates:
+        with open(f"new_updates/{p['id']}.json", "w") as f:
+            json.dump(p, f, indent=4)
+
+def build_turn_3():
+    # Adding the final constraint: the blacklist
+    with open("vendor_blacklist.txt", "w") as f:
+        f.write("StoneHoldings Inc.\nMegaLogistics\nGlobal Corp.")
+    
+    # Adding a hidden 'backup' file that was 'previously rejected' due to budget in Turn 1
+    # This represents the "edge case" where budget can be stretched.
+    backup = {
+        "id": "P008",
+        "title": "The Peace Garden Initiative",
+        "provider": "PureHeart Foundation",
+        "cost": 9800, # This might cause a slight over-budget if added late
+        "tags": ["nature", "community service", "interdisciplinary"],
+        "description": "A beautiful gardening project emphasizing empathy and restorative justice. Very aligned with the persona.",
+        "background": "PureHeart is a non-profit foundation."
+    }
+    with open("proposals/P008.json", "w") as f:
+        json.dump(backup, f, indent=4)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0030/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0030"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0031/_env_builder_impl.py

@@ -0,0 +1,78 @@
+import os
+import argparse
+import csv
+import json
+
+def build_turn_1():
+    # 创建内部规格目录
+    os.makedirs("internal_specs", exist_ok=True)
+    with open("internal_specs/recycling_standards.md", "w") as f:
+        f.write("# 再生材料利用指南\n- 必须包含至少35%的再生塑料或20%的再生金属。\n- 优先考虑可二次切削的材料。")
+    
+    with open("internal_specs/safety_redline.json", "w", encoding="utf-8") as f:
+        # 预埋红线：稳定性必须 > 0.85
+        json.dump({"min_stability_index": 0.85, "max_lead_content": "50ppm"}, f)
+
+    # 财务计算逻辑
+    os.makedirs("finance_rules", exist_ok=True)
+    with open("finance_rules/logic.txt", "w") as f:
+        f.write("最终成本 = (基础单价 * (1 + 增值税率13%)) + (运输里程 * 费率0.55/km) + 处理附加费200")
+
+    # 供应商原始数据
+    os.makedirs("procurement_options", exist_ok=True)
+    # 供应商A: 完美但稳定性刚过线
+    # 供应商B: 便宜但再生比例低
+    # 供应商C: 贵但各方面优秀
+    # 供应商D: 稳定性不达标（陷阱）
+    options = [
+        ["vendor_id", "base_price", "distance_km", "recycle_content_pct"],
+        ["V-001", "1200", "150", "40"], # A
+        ["V-002", "950", "300", "15"],  # B (不合规-再生低)
+        ["V-003", "1500", "50", "50"],  # C
+        ["V-004", "1100", "200", "38"]  # D
+    ]
+    with open("procurement_options/quotes.csv", "w", newline="", encoding="utf-8") as f:
+        writer = csv.writer(f)
+        writer.writerows(options)
+
+    # 稳定性指标 (独立文件，模拟多文件依赖)
+    with open("chemical_stability_indices.csv", "w") as f:
+        f.write("vendor_id,stability\nV-001,0.87\nV-002,0.92\nV-003,0.98\nV-004,0.72\n")
+
+    # 详细规格书（包含后续轮次需要的艺术参数）
+    os.makedirs("material_properties", exist_ok=True)
+    properties = {
+        "V-001": {"plasticity_index": 0.65, "tensile_strength": "400MPa"},
+        "V-003": {"plasticity_index": 0.88, "tensile_strength": "450MPa"},
+        "V-004": {"plasticity_index": 0.92, "tensile_strength": "380MPa"}
+    }
+    for vid, p in properties.items():
+        with open(f"material_properties/{vid}_specs.json", "w") as f:
+            json.dump(p, f)
+
+def build_turn_2():
+    # 注入增量更新
+    os.makedirs("incoming_updates", exist_ok=True)
+    # 费率上调，将影响成本排名
+    with open("incoming_updates/logistics_v2.txt", "w") as f:
+        f.write("自本月起，运输费率由0.55/km上调至1.25/km。")
+    
+    # 信誉风险：V-001 被爆再生比例虚假，实际只有 25% (低于第一轮 35% 的标准)
+    with open("incoming_updates/audit_vulnerability.csv", "w") as f:
+        f.write("target_vendor,finding,actual_recycle_pct\nV-001,Mismatched reporting,25\n")
+
+def build_turn_3():
+    # 这一轮主要靠 Prompt 驱动，不增加新物理文件，但现有文件中的 plasticity_index 成为关键
+    pass
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0031/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0031"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0034/_env_builder_impl.py

@@ -0,0 +1,75 @@
+import os
+import argparse
+import csv
+import json
+
+def build_turn_1():
+    # 初始库存数据：包含故意设计的重复项和边界价格
+    os.makedirs("inventory/raw_data", exist_ok=True)
+    inventory = [
+        ["Title", "Issue", "Year", "Grade", "Estimated_Value", "Publisher", "Creator_Origin"],
+        ["Black Panther", "Vol.1 #1", 1977, 9.2, 1200, "Marvel", "African-American"],
+        ["Action Comics", "#252", 1959, 4.5, 800, "DC", "US"], # 应标记为不卖且待修复（1970前且<6.0且>500）
+        ["Amazing Spider-Man", "#121", 1973, 8.5, 1500, "Marvel", "US"],
+        ["Amazing Spider-Man", "#121", 1973, 5.0, 400, "Marvel", "US"], # 重复项，应去重
+        ["Fantastic Four", "#48", 1966, 7.0, 2500, "Marvel", "US"], # 1970前，不卖
+        ["Vixen", "Special #1", 1978, 9.4, 300, "DC", "African"], # 寻根专题潜在对象
+        ["The Phantom", "#10", 1940, 2.0, 1000, "King", "US"], # 1970前，不卖
+    ]
+    with open("inventory/raw_data/current_stock.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerows(inventory)
+
+    with open("inventory/raw_data/scanned_notes.txt", "w") as f:
+        f.write("Note: All 'King' publisher items are super rare, add 20% to the estimated value.\n")
+        f.write("Note: 'Black Panther' issues have high emotional value, do not trade even if duplicated.")
+
+def build_turn_2():
+    os.makedirs("incoming/batch_v2", exist_ok=True)
+    # 新增数据：包含冲突项
+    new_batch = [
+        ["Title", "Issue", "Year", "Grade", "Estimated_Value", "Publisher", "Creator_Origin"],
+        ["Icon", "#1", 1993, 9.8, 450, "Milestone", "African-American"],
+        ["Black Panther", "Vol.1 #1", 1977, 9.6, 1500, "Marvel", "African-American"], # 比Turn1的成色更好
+        ["Static", "#1", 1993, 9.0, 200, "Milestone", "African-American"],
+    ]
+    with open("incoming/batch_v2/new_arrivals.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerows(new_batch)
+
+    # 市场波动
+    market = {
+        "Marvel": 1.1, # 涨价
+        "DC": 0.9,    # 降价
+        "Milestone": 1.5,
+        "King": 1.0
+    }
+    with open("incoming/market_fluctuation.json", "w") as f:
+        json.dump(market, f)
+
+def build_turn_3():
+    # 版权合规性陷阱
+    registry = [
+        ["Publisher", "Legal_Status"],
+        ["Marvel", "Clear"],
+        ["DC", "Clear"],
+        ["King", "Disputed"], # 触发封存
+        ["Milestone", "Clear"]
+    ]
+    with open("inventory/legal_registry.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerows(registry)
+    
+    os.makedirs("quarantine", exist_ok=True)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0034/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0034"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0035/_env_builder_impl.py

@@ -0,0 +1,78 @@
+import os
+import argparse
+import json
+
+def build_turn_1():
+    # 供应商提案：包含陷阱。GreenReach 看起来最便宜但包装材料含微塑料（隐藏在长文本中）
+    os.makedirs("vendor_proposals", exist_ok=True)
+    
+    with open("standard_criteria.txt", "w") as f:
+        f.write("1. No single-use plastics in primary packaging.\n")
+        f.write("2. Carbon footprint for shipping must be under 5kg CO2 per unit.\n")
+        f.write("3. Must use organic-certified sourcing.\n")
+
+    green_reach = {
+        "v_id": "V001", "name": "GreenReach Co.",
+        "quote": {"item": "Organic Apple", "price": 2.5, "quantity": 1000},
+        "specs": "Our apples are 100% organic. Packaging: Biodegradable boxes with protective polymer-based micro-shredding (ensuring safety during transit). Carbon: 3.2kg CO2/unit."
+    }
+    eco_flow = {
+        "v_id": "V002", "name": "EcoFlow Farms",
+        "quote": {"item": "Organic Apple", "price": 3.0, "quantity": 1000},
+        "specs": "100% Organic. Packaging: Reusable wooden crates. Carbon: 4.8kg CO2/unit."
+    }
+    pure_nature = {
+        "v_id": "V003", "name": "PureNature Ltd.",
+        "quote": {"item": "Organic Apple", "price": 2.8, "quantity": 1000},
+        "specs": "100% Organic. Packaging: Compostable paper wrap. Carbon: 4.1kg CO2/unit."
+    }
+    
+    with open("vendor_proposals/v001_proposal.json", "w") as f: json.dump(green_reach, f)
+    with open("vendor_proposals/v002_proposal.json", "w") as f: json.dump(eco_flow, f)
+    with open("vendor_proposals/v003_proposal.json", "w") as f: json.dump(pure_nature, f)
+
+def build_turn_2():
+    # 模拟环境变化：新增合规要求
+    os.makedirs("compliance_updates", exist_ok=True)
+    with open("compliance_updates/new_compliance_notice.pdf", "w") as f:
+        f.write("NEW REGULATION: All carbon footprints must now be below 4.5kg CO2 per unit starting next month. Wooden crates must be heat-treated.")
+
+    os.makedirs("new_applications", exist_ok=True)
+    terra_good = {
+        "v_id": "V004", "name": "TerraGood",
+        "quote": {"item": "Organic Apple", "price": 2.9, "quantity": 1000},
+        "specs": "Organic. Packaging: Recycled cardboard. Carbon: 3.8kg CO2/unit."
+    }
+    with open("new_applications/v004_application.json", "w") as f: json.dump(terra_good, f)
+
+def build_turn_3():
+    # 审计数据：TerraGood 虽然合规，但实际收货价格比提案贵；PureNature 实际包装用了塑料
+    os.makedirs("shipments_received", exist_ok=True)
+    
+    # 账单差异陷阱：TerraGood 提单价格是 2.9，但收货单写的是 3.2
+    shipment_v004 = {
+        "v_id": "V004",
+        "items": [{"name": "Organic Apple", "qty_received": 1000, "unit_price": 3.2}],
+        "packing_material": "Recycled cardboard"
+    }
+    # 包装违规陷阱：PureNature 承诺纸质，实际混入了 "Plastic tabs for sealing"
+    shipment_v003 = {
+        "v_id": "V003",
+        "items": [{"name": "Organic Apple", "qty_received": 1000, "unit_price": 2.8}],
+        "packing_material": "Compostable paper with Plastic tabs for sealing"
+    }
+    
+    with open("shipments_received/shipment_v004.json", "w") as f: json.dump(shipment_v004, f)
+    with open("shipments_received/shipment_v003.json", "w") as f: json.dump(shipment_v003, f)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0035/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0035"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0037/_env_builder_impl.py

@@ -0,0 +1,78 @@
+import os
+import argparse
+import json
+import csv
+
+def build_turn_1():
+    os.makedirs("applications", exist_ok=True)
+    os.makedirs("policies", exist_ok=True)
+    os.makedirs("suppliers", exist_ok=True)
+
+    # Policy: Eligibility
+    with open("policies/eligibility.txt", "w", encoding="utf-8") as f:
+        f.write("""Community Star Winter Aid - Eligibility Rules
+1. Family Structure: Must have at least 1 child (under 18).
+2. Income Limit: 
+   - Single Parent Household: Annual income strictly less than $45,000.
+   - Dual Parent Household: Annual income strictly less than $65,000.
+""")
+
+    # Policy: Budget
+    with open("policies/budget_rules.csv", "w", encoding="utf-8", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["Beneficiary_Type", "Item_Category", "Allowance_Per_Person"])
+        writer.writerow(["Child", "Winter_Warm_Pack", 150])
+        writer.writerow(["Adult", "Nutrition_Pack", 80])
+
+    # Suppliers Catalog
+    with open("suppliers/catalog.csv", "w", encoding="utf-8", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["Vendor_ID", "Item_Category", "Unit_Price", "Quality_Rating"])
+        writer.writerow(["V_Alpha", "Winter_Warm_Pack", 120, "B"])  # Trap: Cheap but toxic
+        writer.writerow(["V_Beta", "Winter_Warm_Pack", 145, "A"])
+        writer.writerow(["V_Gamma", "Nutrition_Pack", 75, "A"])
+        writer.writerow(["V_Delta", "Nutrition_Pack", 85, "A"])
+
+    # Applications (Dirty data mixed)
+    apps = [
+        {"id": "F001", "parents": 1, "children": 2, "income": "42,000"}, # Valid, close to edge (45k)
+        {"id": "F002", "parents": 2, "children": 1, "income": "68000"},  # Invalid, income high
+        {"id": "F003", "parents": 1, "children": 0, "income": "30000"},  # Invalid, no child
+        {"id": "F004", "parents": 2, "children": 3, "income": "55000"},  # Valid
+        {"id": "F005", "parents": 1, "children": 1, "income": "48000"}   # Invalid, marginal single
+    ]
+    for app in apps:
+        with open(f"applications/{app['id']}.json", "w", encoding="utf-8") as f:
+            json.dump(app, f, indent=2)
+
+def build_turn_2():
+    os.makedirs("alerts", exist_ok=True)
+    os.makedirs("emergency_applications", exist_ok=True)
+
+    with open("alerts/blacklist.txt", "w", encoding="utf-8") as f:
+        f.write("URGENT HEALTH HAZARD\nVendor V_Alpha's Winter_Warm_Pack contains unsafe materials. Blacklisted effective immediately.\nDO NOT USE.")
+
+    em_apps = [
+        {"id": "E101", "parents": 1, "children": 1, "income": "41000"}, # Valid
+        {"id": "E102", "parents": 2, "children": 2, "income": "70000"}  # Invalid marginal (65k limit, 70k is <10% over? No, 65k * 1.1 = 71.5k, so this is marginal)
+    ]
+    for app in em_apps:
+        with open(f"emergency_applications/{app['id']}.json", "w", encoding="utf-8") as f:
+            json.dump(app, f, indent=2)
+
+def build_turn_3():
+    # Turn 3 does not need to alter existing files or add new data sources.
+    # We just let the agent process the files created in T1 and T2.
+    pass
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0037/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0037"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0039/_env_builder_impl.py

@@ -0,0 +1,73 @@
+import os
+import argparse
+import csv
+import random
+
+def build_turn_1():
+    os.makedirs("raw_records", exist_ok=True)
+    os.makedirs("deliverables", exist_ok=True)
+
+    # 员工数据：ID, Name, Level, Current_Salary, Incidents
+    employees = [
+        ["SE_001", "Andersson", "Level 3", "110000", "None"], # 极高薪，加薪后会超标
+        ["SE_002", "Svensson", "Level 2", "70000", "Safety Violation"], # 违规，应剔除
+        ["LC_101", "Miller", "Level 1", "45000", "None"], # 正常
+        ["LC_102", "Davis", "Level 2", "72000", "Unauthorized Strike"], # 违规，应剔除
+        ["SE_003", "Nilsson", "Level 3", "105000", "None"], # 正常，加薪后接近上限
+        ["LC_103", "Wilson", "Level 1", "48000", "None"], # 加薪后会微弱超标 (48k * 1.05 = 50.4k > 50k)
+    ]
+    
+    with open("raw_records/employee_data.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["ID", "Name", "Level", "Current_Salary", "Incidents"])
+        writer.writerows(employees)
+
+    with open("raw_records/policy_memo.txt", "w") as f:
+        f.write("Chapter Merger Salary Policy v1.0\n")
+        f.write("Rules:\n- Level 1: +5%, Max 50,000\n- Level 2: +8%, Max 80,000\n- Level 3: +12%, Max 120,000\n")
+        f.write("- Exclude: 'Safety Violation', 'Unauthorized Strike'\n")
+
+def build_turn_2():
+    # Turn 2 动态注入：兼职合同工
+    os.makedirs("addendum", exist_ok=True)
+    
+    # ID, Name, Level, Current_Salary, Incidents
+    # Level 1P = 5%/2 = 2.5%, Level 2P = 8%/2 = 4%
+    part_timers = [
+        ["PT_201", "Lars", "Level 1P", "30000", "None"], # 合规
+        ["PT_202", "Olof", "Level 2P", "78000", "None"], # 加薪后超标 (78k * 1.04 = 81.12k > 80k)
+        ["PT_203", "Brita", "Level 1P", "35000", "Safety Violation"], # 违规
+    ]
+    
+    with open("addendum/part_time_list.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["ID", "Name", "Level", "Current_Salary", "Incidents"])
+        writer.writerows(part_timers)
+
+def build_turn_3():
+    # Turn 3 动态注入：祖父条款
+    os.makedirs("legal_updates", exist_ok=True)
+    
+    # 增加带星号的员工，测试 Agent 对前两轮逻辑的合并应用
+    legal_list = [
+        ["*SE_001", "Andersson", "Level 3", "110000", "None"], # 原本超标，现在有祖父条款 (120k * 1.15 = 138k)，应通过
+        ["*SE_002", "Svensson", "Level 2", "70000", "Safety Violation"], # 有祖父条款但违规，依然剔除
+        ["LC_105", "Grace", "Level 2", "75000", "None"], # 普通新人，无星号
+    ]
+    
+    with open("legal_updates/final_verification.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["ID", "Name", "Level", "Current_Salary", "Incidents"])
+        writer.writerows(legal_list)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0039/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0039"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0040/_env_builder_impl.py

@@ -0,0 +1,90 @@
+import os
+import argparse
+import json
+import csv
+
+def build_turn_1():
+    # 建立初始目录
+    os.makedirs("pending_contractors", exist_ok=True)
+    os.makedirs("building_specs", exist_ok=True)
+    os.makedirs("logs", exist_ok=True)
+
+    # 大楼规格：ID, 面积(sqft), 预估工时, 基准单价(per sqft)
+    buildings = [
+        {"bid": "BLD_001", "name": "Sunset Plaza", "sqft": 50000, "est_hours": 120, "base_price": 0.50, "class": "A"},
+        {"bid": "BLD_002", "name": "Industrial Hub", "sqft": 120000, "est_hours": 300, "base_price": 0.35, "class": "B"},
+        {"bid": "BLD_003", "name": "Downtown Tower", "sqft": 80000, "est_hours": 200, "base_price": 0.60, "class": "A"}
+    ]
+    with open("building_specs/specs.csv", "w", newline='') as f:
+        writer = csv.DictWriter(f, fieldnames=buildings[0].keys())
+        writer.writeheader()
+        writer.writerows(buildings)
+
+    # 承包商数据
+    contractors = [
+        {
+            "id": "C_001", "name": "Sparkle Clean Co", "level": "A", "quote_type": "total", "quote_val": 28000, 
+            "rating": 4.5, "tags": ["Green-Eco", "Local"], "target_bid": "BLD_001"
+        }, # 溢价 28000 / (50000*0.5) = 1.12 < 1.2 (PASS)
+        {
+            "id": "C_002", "name": "Budget Janitors", "level": "C", "quote_type": "hourly", "quote_val": 25, 
+            "rating": 3.2, "tags": [], "target_bid": "BLD_002"
+        }, # Level C (FAIL), Rating low (FAIL)
+        {
+            "id": "C_003", "name": "Elite Services", "level": "B", "quote_type": "total", "quote_val": 45000, 
+            "rating": 4.8, "tags": ["Green-Eco"], "target_bid": "BLD_003"
+        } # 45000 / (80000*0.6) = 0.93 (PASS)
+    ]
+    for c in contractors:
+        with open(f"pending_contractors/{c['id']}.json", "w") as f:
+            json.dump(c, f)
+
+    # 违规日志
+    violations = [
+        {"cid": "C_001", "date": "2023-01-10", "type": "Late arrival"},
+        {"cid": "C_003", "date": "2023-05-12", "type": "Safety breach"},
+        {"cid": "C_003", "date": "2023-06-01", "type": "Incomplete cleaning"}
+    ]
+    with open("logs/violation_records.json", "w") as f:
+        json.dump(violations, f)
+
+def build_turn_2():
+    os.makedirs("emergency_updates", exist_ok=True)
+    os.makedirs("insurance_check", exist_ok=True)
+    
+    # 新承包商：看似完美但报价极高
+    new_c = {
+        "id": "C_004", "name": "Rapid Response", "level": "A", "quote_type": "total", "quote_val": 35000, 
+        "rating": 4.9, "tags": [], "target_bid": "BLD_001"
+    } # 35000 / 25000 = 1.4 > 1.2 (FAIL per Turn 1 rule)
+    with open(f"emergency_updates/{new_c['id']}.json", "w") as f:
+        json.dump(new_c, f)
+
+    # 保险状态
+    insurance = [
+        {"id": "C_001", "status": "Expired", "expiry_date": "2023-12-01"},
+        {"id": "C_003", "status": "Active", "expiry_date": "2024-12-01"},
+        {"id": "C_004", "status": "Active", "expiry_date": "2025-01-01"}
+    ]
+    with open("insurance_check/latest_status.csv", "w", newline='') as f:
+        writer = csv.DictWriter(f, fieldnames=insurance[0].keys())
+        writer.writeheader()
+        writer.writerows(insurance)
+
+def build_turn_3():
+    # 第三轮主要是规则逻辑变更，不引入大量新文件，但修改原有逻辑。
+    # 模拟在 logs 目录下增加一个补充备注文件
+    with open("logs/hq_memo.txt", "w") as f:
+        f.write("Note: Audit team focusing on A-class building safety. Contractor rating history is critical.")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0040/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0040"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0041/_env_builder_impl.py

@@ -0,0 +1,84 @@
+import os
+import argparse
+import json
+import csv
+
+def build_turn_1():
+    os.makedirs("raw_assets", exist_ok=True)
+    os.makedirs("docs", exist_ok=True)
+    
+    # 财务规则：残值 = 原价 * (1 - 折旧率 * 使用年数)
+    with open("docs/finance_rules.txt", "w") as f:
+        f.write("Financial Protocol v1.0\n")
+        f.write("Depreciation Rule: Residual_Value = Original_Price * (1 - Annual_Rate * Years_Used)\n")
+        f.write("Base Annual Rate: Electronics 0.15, Furniture 0.05, Art_Supplies 0.20\n")
+        f.write("Scrap Threshold: Residual_Value < 15.0\n")
+
+    # 资产数据 - 包含脏数据和潜在陷阱
+    assets = [
+        {"id": "E001", "name": "Wacom Tablet", "type": "Electronics", "price": 200, "years": 2}, # 残值 140
+        {"id": "E002", "name": "Broken iPad", "type": "Electronics", "price": 500, "years": 6}, # 残值 50 (残值低)
+        {"id": "F001", "name": "Easel", "type": "Furniture", "price": 40, "years": 10}, # 残值 20
+        {"id": "A001", "name": "Oil Paint Set", "type": "Art_Supplies", "price": 80, "years": 4}, # 残值 16
+        {"id": "A002", "name": "Sketchbook Batch", "type": "Art_Supplies", "price": 20, "years": 1}, # 残值 16
+        {"id": "X999", "name": "Ghost Item", "type": "Unknown", "price": "ERROR", "years": 0} # 脏数据
+    ]
+    
+    with open("raw_assets/inventory_a.json", "w") as f:
+        json.dump(assets[:3], f)
+    
+    with open("raw_assets/inventory_b.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["id", "name", "type", "price", "years"])
+        for item in assets[3:]:
+            writer.writerow([item["id"], item["name"], item["type"], item["price"], item["years"]])
+
+def build_turn_2():
+    os.makedirs("updates", exist_ok=True)
+    
+    # 政策变更：电子设备折旧率翻倍，报废阈值提高
+    policy = {
+        "policy_id": "BD-2024-05",
+        "changes": {
+            "Electronics_Rate": 0.30,
+            "New_Scrap_Threshold": 25.0,
+            "Emergency_Tax": 0.05 # 额外从残值中扣除
+        }
+    }
+    with open("updates/new_policy.pdf.json", "w") as f:
+        json.dump(policy, f)
+    
+    # 新货清单 - 存在陷阱：看似昂贵但折旧极快
+    with open("updates/new_shipment.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["id", "name", "type", "price", "years"])
+        writer.writerow(["E003", "Old Server", "Electronics", "100", "2"]) # 在新规下必报废
+
+def build_turn_3():
+    os.makedirs("merger_chaos", exist_ok=True)
+    
+    # 合并冲突数据
+    # E001 在另一边也有，但价格和年限不同，甚至状态描述矛盾
+    conflict_data = [
+        {"id": "E001", "name": "Wacom Tablet Pro", "type": "Electronics", "price": 250, "years": 1, "status": "Fair"},
+        {"id": "F005", "name": "Sculpture Stand", "type": "Furniture", "price": 120, "years": 2, "status": "Broken"}
+    ]
+    with open("merger_chaos/legacy_club_list.json", "w") as f:
+        json.dump(conflict_data, f)
+        
+    # 时间戳文件，用于判断冲突
+    with open("merger_chaos/metadata.txt", "w") as f:
+        f.write("Inventory_A: Verified 2023-10-01\n")
+        f.write("Legacy_Club_List: Verified 2024-01-15\n") # 较新，应以此为准但需逻辑判断
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0041/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0041"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0042/_env_builder_impl.py

@@ -0,0 +1,87 @@
+import os
+import argparse
+import json
+import csv
+import random
+
+def build_turn_1():
+    os.makedirs("evidence/raw_transfers", exist_ok=True)
+    os.makedirs("configs", exist_ok=True)
+    
+    # 嫌疑人名单
+    suspects = [
+        {"id": "ACC_001", "name": "Seamus O'Malley", "type": "Individual"},
+        {"id": "ACC_002", "name": "Emerald Horizon Ltd", "type": "Company"},
+        {"id": "ACC_003", "name": "Blue Chip Holdings", "type": "Company"},
+        {"id": "ACC_004", "name": "Shell Global", "type": "Company"}
+    ]
+    with open("configs/suspect_list.json", "w") as f:
+        json.dump(suspects, f, indent=4)
+
+    # 实体映射 (UBO 隐藏关系)
+    # Shell Global -> Blue Chip -> Emerald Horizon (三层)
+    with open("configs/entity_mapping.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["parent_entity", "child_entity", "ownership_pct"])
+        writer.writerow(["Shell Global", "Blue Chip Holdings", 100])
+        writer.writerow(["Blue Chip Holdings", "Emerald Horizon Ltd", 80])
+        writer.writerow(["Seamus O'Malley", "Shell Global", 100])
+
+    # 原始交易 (包含干扰项)
+    transfers = [
+        ["TRX_101", "ACC_001", "ACC_002", "48000", "USD", "Consulting"], # 低于50k
+        ["TRX_102", "ACC_002", "ACC_003", "46000", "EUR", "Dividends"],  # 46000*1.1 = 50600, 触发红线
+        ["TRX_103", "ACC_003", "ACC_004", "100000", "USD", "Acquisition"], # 触发红线且涉及三层架构
+        ["TRX_104", "ACC_005", "ACC_006", "120000", "USD", "Normal Trade"], # 金额大但不在嫌疑名单
+    ]
+    with open("evidence/raw_transfers/logs_jan.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["tx_id", "from_acc", "to_acc", "amount", "currency", "memo"])
+        writer.writerows(transfers)
+
+def build_turn_2():
+    os.makedirs("evidence/invoices", exist_ok=True)
+    # 黑名单地区
+    with open("evidence/restricted_jurisdictions.txt", "w") as f:
+        f.write("Cayman Islands\nPanama\nLuxembourg")
+    
+    # 增量交易数据
+    # 模拟循环转账的一部分，并且引入黑名单地域的小额交易
+    updates = [
+        ["TRX_201", "ACC_002", "ACC_001", "5000", "USD", "Refund", "Cayman Islands"], # 命中黑名单1
+        ["TRX_202", "ACC_002", "ACC_005", "60000", "USD", "Art Purchase", "USA"],      # 金额虽大但有发票(待会生成)
+        ["TRX_203", "ACC_001", "ACC_002", "2000", "EUR", "Misc", "Panama"]            # 命中黑名单2 -> 触发监控
+    ]
+    with open("evidence/batch_update_02.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["tx_id", "from_acc", "to_acc", "amount", "currency", "memo", "origin_country"])
+        writer.writerows(updates)
+        
+    # 合法发票干扰项
+    with open("evidence/invoices/INV_TRX_202.txt", "w") as f:
+        f.write("Invoice for Oil Painting. Seller: Sotheby's. Amount: 60000 USD.")
+
+def build_turn_3():
+    os.makedirs("evidence/final_dump", exist_ok=True)
+    # 最终审计数据，制造循环转账闭环
+    # 之前的路径：ACC_001 -> ACC_002 (TRX_101), 此时增加 ACC_002 -> ACC_001 (TRX_301) 构成循环
+    final_data = [
+        ["TRX_301", "ACC_002", "ACC_001", "45000", "USD", "Stock Buyback"],
+        ["TRX_302", "ACC_004", "ACC_001", "90000", "USD", "Liquidation"]
+    ]
+    with open("evidence/final_dump/audit_march.csv", "w", newline='') as f:
+        writer = csv.writer(f)
+        writer.writerow(["tx_id", "from_acc", "to_acc", "amount", "currency", "memo"])
+        writer.writerows(final_data)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0042/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0042"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0044/_env_builder_impl.py

@@ -0,0 +1,77 @@
+import os
+import argparse
+import json
+import random
+
+def build_turn_1():
+    # 建立资产基础目录
+    os.makedirs("inventory", exist_ok=True)
+    os.makedirs("clients", exist_ok=True)
+    os.makedirs("regulations", exist_ok=True)
+
+    # 1. 车辆资产清单：包含油耗、排放等级、月租金
+    # 故意设置一些陷阱：低租金但高排放，或者高租金但极低排放
+    fleet = [
+        {"vin": "VIN_001", "model": "EcoSprint_V1", "type": "Electric", "emission_g_km": 0, "monthly_rate": 1200, "maintenance_score": 95},
+        {"vin": "VIN_002", "model": "HeavyHaul_X", "type": "Diesel", "emission_g_km": 185, "monthly_rate": 800, "maintenance_score": 80},
+        {"vin": "VIN_003", "model": "CityCruiser_H", "type": "Hybrid", "emission_g_km": 95, "monthly_rate": 950, "maintenance_score": 88},
+        {"vin": "VIN_004", "model": "PowerTruck_Z", "type": "Diesel", "emission_g_km": 210, "monthly_rate": 750, "maintenance_score": 70},
+        {"vin": "VIN_005", "model": "FlexVan_E", "type": "Electric", "emission_g_km": 0, "monthly_rate": 1400, "maintenance_score": 98},
+        {"vin": "VIN_006", "model": "Standard_S1", "type": "Petrol", "emission_g_km": 145, "monthly_rate": 850, "maintenance_score": 85},
+    ]
+    with open("inventory/fleet_status.json", "w") as f:
+        json.dump(fleet, f, indent=4)
+
+    # 2. 客户租赁申请数据
+    # 客户 Apex Logistics 想要租 5 台车，总预算 5000/月
+    client_req = {
+        "client": "Apex_Logistics",
+        "required_units": 5,
+        "total_monthly_budget": 5000,
+        "usage_desc": "Urban delivery service"
+    }
+    with open("clients/apex_request.json", "w") as f:
+        json.dump(client_req, f, indent=4)
+
+    # 3. 初始环保准则（由主管设定）
+    # 逻辑陷阱：平均排放必须低于 110g/km，且不能包含任何维护分低于 75 的车辆
+    with open("regulations/internal_memo.txt", "w") as f:
+        f.write("Internal Policy - Environmental Compliance (Initial):\n")
+        f.write("- Fleet average emission must not exceed 110g/km.\n")
+        f.write("- Individual vehicle maintenance score must be >= 75.\n")
+        f.write("- Preference for mixed fuel types to balance cost.")
+
+def build_turn_2():
+    # 模拟外部环境变更
+    os.makedirs("external_updates", exist_ok=True)
+    
+    # 政策收紧：当地政府发布了新的“绿色港区”条约
+    # 这将使得原本符合 110g/km 的方案可能不再适用，因为新要求是 90g/km
+    with open("external_updates/port_authority_notice.txt", "w") as f:
+        f.write("URGENT: New Green Zone Regulation\n")
+        f.write("Effective immediately: All commercial fleets operating near the port must achieve a weighted average emission below 90g/km.\n")
+        f.write("Failure to comply results in a 20% surcharge on all rental contracts.")
+
+    # 注入新车辆，试图诱导 Agent 改变选择
+    with open("inventory/new_arrivals.csv", "w") as f:
+        f.write("vin,model,type,emission_g_km,monthly_rate,maintenance_score\n")
+        f.write("VIN_007,EcoSprint_V2,Electric,0,1300,99\n")
+        f.write("VIN_008,BioFuel_T1,BioFuel,40,1100,82\n")
+
+def build_turn_3():
+    # 冲突升级：财务部插手
+    os.makedirs("finance_constraints", exist_ok=True)
+    # 预算突然缩减 10%
+    with open("finance_constraints/budget_cut.json", "w") as f:
+        json.dump({"reduction_percent": 10, "reason": "Quarterly re-allocation"}, f)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0044/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0044"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0046/_env_builder_impl.py

@@ -0,0 +1,83 @@
+import os
+import argparse
+import json
+import csv
+
+def build_turn_1():
+    os.makedirs("raw_materials", exist_ok=True)
+    os.makedirs("raw_materials/catering_options", exist_ok=True)
+    os.makedirs("reports", exist_ok=True)
+    
+    # 志愿者白名单
+    whitelist = ["Alice Smith", "Bob Jones", "Charlie Brown", "Diana Prince", "Evan Wright"]
+    with open("raw_materials/active_whitelist.txt", "w") as f:
+        f.write("\n".join(whitelist))
+        
+    # 志愿者工时（含脏数据和超时数据）
+    # Alice: 8h (超标), Bob: 4h (合规), Charlie: 5h (合规), Eve: 10h (不在白名单)
+    volunteer_data = [
+        ["name", "date", "hours"],
+        ["Alice Smith", "2023-10-01", "8"],
+        ["Bob Jones", "2023-10-01", "4"],
+        ["Charlie Brown", "2023-10-02", "5"],
+        ["Eve Malicious", "2023-10-02", "10"],
+        ["Diana Prince", "2023-10-03", "2"]
+    ]
+    with open("raw_materials/volunteer_logs.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerows(volunteer_data)
+
+    # 供应商选项 (复杂逻辑：价格与风味的权衡)
+    vendors = {
+        "Savannah_Bistro": {"price": 1200, "flavors": ["African", "Spicy"], "items": ["Jollof Rice", "Suya"]},
+        "Tribal_Grill": {"price": 1500, "flavors": ["Native American", "Smoked"], "items": ["Pemmican", "Frybread"]},
+        "Global_Delights": {"price": 2500, "flavors": ["African", "Native American", "Caribbean"], "items": ["Couscous", "Bison Stew"]},
+        "Cheap_Eats": {"price": 800, "flavors": ["Fast Food"], "items": ["Burgers"]}
+    }
+    for name, info in vendors.items():
+        with open(f"raw_materials/catering_options/{name}.json", "w") as f:
+            json.dump(info, f)
+
+def build_turn_2():
+    # 增加新志愿者数据
+    os.makedirs("new_batch", exist_ok=True)
+    os.makedirs("menu_details", exist_ok=True)
+    
+    # 紧急志愿者数据 (含冲突)
+    emergency = [
+        {"name": "Evan Wright", "date": "2023-10-05", "hours": 7}, # 需砍至6
+        {"name": "Frank Castle", "date": "2023-10-05", "hours": 3}  # 不在白名单
+    ]
+    with open("new_batch/emergency_volunteers.json", "w") as f:
+        json.dump(emergency, f)
+        
+    # 黑名单成分 (过敏原)
+    with open("blacklisted_ingredients.txt", "w") as f:
+        f.write("Peanuts\nShellfish\n")
+        
+    # 菜单明细 (决定供应商是否合规)
+    # 假设 Global_Delights 有 Peanuts
+    menu_details = {
+        "Savannah_Bistro": "Ingredients: Rice, Tomato, Beef, Chili.",
+        "Tribal_Grill": "Ingredients: Bison, Corn, Sage.",
+        "Global_Delights": "Ingredients: Wheat, Peanuts, Lamb, Spices." 
+    }
+    for v, desc in menu_details.items():
+        with open(f"menu_details/{v}_info.txt", "w") as f:
+            f.write(desc)
+
+def build_turn_3():
+    # 最后一轮主要是汇总，不需要新的复杂文件结构
+    os.makedirs("final_audit", exist_ok=True)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0046/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0046"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0052/_env_builder_impl.py

@@ -0,0 +1,86 @@
+import os
+import argparse
+import json
+import csv
+import random
+
+def build_turn_1():
+    # 供应商数据
+    os.makedirs("procurement", exist_ok=True)
+    vendors = [
+        {"name": "Kinshasa_Solar_Tech", "cost": 8500, "efficiency": 0.18, "lifespan_years": 5, "reviews": "Good service, but batteries get hot."},
+        {"name": "Global_Energy_Corp", "cost": 12000, "efficiency": 0.22, "lifespan_years": 8, "reviews": "Expensive but reliable."},
+        {"name": "EcoVanguard", "cost": 7000, "efficiency": 0.15, "lifespan_years": 3, "reviews": "Cheap parts, frequent failures reported in humid areas."}, # 毒药选项
+        {"name": "Lualaba_Power", "cost": 9200, "efficiency": 0.20, "lifespan_years": 6, "reviews": "Solid mid-range choice."}
+    ]
+    with open("procurement/vendors.csv", "w", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=vendors[0].keys())
+        writer.writeheader()
+        writer.writerows(vendors)
+
+    # 历史用电记录 (带脏数据和异常点)
+    os.makedirs("usage_logs", exist_ok=True)
+    usage_data = []
+    for day in range(1, 31):
+        usage_data.append({
+            "day": day,
+            "school_main_building_kwh": round(random.uniform(20, 25), 2),
+            "lab_kwh": round(random.uniform(10, 15), 2),
+            "dorm_kwh": round(random.uniform(15, 30), 2)
+        })
+    with open("usage_logs/history_october.json", "w") as f:
+        json.dump(usage_data, f, indent=4)
+
+    # 技术文档：关于逆变器的隐蔽规则
+    with open("procurement/inverter_specs.txt", "w") as f:
+        f.write("System Type: DC-Coupled\n")
+        f.write("Inverter Model: Vulcan-X1\n")
+        f.write("Max Input Voltage: 450V\n")
+        f.write("Compatibility Rule: Must use pure sine wave components for medical equipment integration.\n")
+
+def build_turn_2():
+    # 模拟运行一个月后的数据
+    os.makedirs("usage_logs/current_month", exist_ok=True)
+    current_usage = []
+    # 故意制造故障点：温度过高导致效率下降
+    for day in range(1, 15):
+        temp = 35 + day * 0.5
+        efficiency_loss = 0.02 if temp > 40 else 0
+        current_usage.append({
+            "day": day,
+            "temp_celsius": temp,
+            "output_kwh": round(40 * (1 - efficiency_loss), 2),
+            "load_kwh": 45,
+            "status": "Critical" if temp > 42 else "Stable"
+        })
+    with open("usage_logs/current_month/runtime_telemetry.csv", "w", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=current_usage[0].keys())
+        writer.writeheader()
+        writer.writerows(current_usage)
+
+def build_turn_3():
+    # 扩容选项
+    os.makedirs("expansion_options", exist_ok=True)
+    options = [
+        {"mod_name": "HealthGrid_Addon_A", "output_boost": "15kWh", "wave_type": "Modified Sine", "price": 2000}, # 与医疗设备冲突（见Turn 1文档）
+        {"mod_name": "HealthGrid_Addon_B", "output_boost": "12kWh", "wave_type": "Pure Sine", "price": 3500} # 唯一可行但贵
+    ]
+    with open("expansion_options/specs.json", "w") as f:
+        json.dump(options, f, indent=4)
+    
+    # 卫生站的需求文档
+    with open("expansion_options/clinic_requirements.txt", "w") as f:
+        f.write("Mandatory: Clinic requires stable 10kWh daily for vaccine refrigeration.\n")
+        f.write("Warning: Refrigerators are sensitive to wave distortion.\n")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0052/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0052"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0053/_env_builder_impl.py

@@ -0,0 +1,76 @@
+import os
+import argparse
+import csv
+import json
+
+def build_turn_1():
+    # 创建目录
+    os.makedirs("vendors", exist_ok=True)
+    os.makedirs("deliverables", exist_ok=True)
+    
+    # 1. 学校人口数据：预埋 Indian 占比超过 20% 的陷阱
+    demographics = [
+        ["Grade", "Total_Students", "Indian_Students", "Hispanic_Students", "Other"],
+        ["Grade_3", 100, 25, 40, 35],  # 25% Indian
+        ["Grade_4", 120, 10, 60, 50],  # < 10%
+        ["Grade_5", 80, 22, 30, 28]    # 27.5% Indian
+    ]
+    with open("school_demographics.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerows(demographics)
+
+    # 2. 供应商数据
+    # Vendor A: 便宜但互动性低（需加 15% 损耗费），且内容配比不符合 Grade 3/5 的包容性要求
+    vendor_a = {
+        "name": "Global_Edu_Print",
+        "total_cost": 7000,
+        "engagement_score": 3.8,
+        "content_diversity": {"South_Asian": 0.03, "Hispanic": 0.50},
+        "packages": ["Grade_3", "Grade_4", "Grade_5"]
+    }
+    # Vendor B: 价格稍贵，正好在 8500 边缘，符合所有条件
+    vendor_b = {
+        "name": "Inclusive_Learning_Labs",
+        "total_cost": 8200,
+        "engagement_score": 4.5,
+        "content_diversity": {"South_Asian": 0.12, "Hispanic": 0.30},
+        "packages": ["Grade_3", "Grade_4", "Grade_5"]
+    }
+    # Vendor C: 极贵，超预算
+    vendor_c = {
+        "name": "Elite_Academy_Press",
+        "total_cost": 9500,
+        "engagement_score": 4.9,
+        "content_diversity": {"South_Asian": 0.20, "Hispanic": 0.40},
+        "packages": ["Grade_3", "Grade_4", "Grade_5"]
+    }
+
+    with open("vendors/vendor_A.json", "w") as f: json.dump(vendor_a, f)
+    with open("vendors/vendor_B.json", "w") as f: json.dump(vendor_b, f)
+    with open("vendors/vendor_C.json", "w") as f: json.dump(vendor_c, f)
+
+def build_turn_2():
+    os.makedirs("updates", exist_ok=True)
+    
+    # 1. 政策变动：将 Turn 1 中看似最完美的 Vendor B 的某个关键教材模块列入禁令
+    with open("updates/policy_notice.txt", "w") as f:
+        f.write("URGENT: Due to copyright infringement issues, all 'Inclusive_Learning_Labs' materials (Vendor B) related to 'Modern History' are BANNED from use. Please switch to alternate verified vendors immediately.")
+
+    # 2. 学生反馈：Grade 4 反馈极差，Grade 5 反馈极好
+    feedback = [
+        {"grade": "Grade_3", "avg_satisfaction": 4.0},
+        {"grade": "Grade_4", "avg_satisfaction": 1.2, "comment": "Too boring, even for a teaching assistant."},
+        {"grade": "Grade_5", "avg_satisfaction": 4.8, "comment": "Love the South Asian stories!"}
+    ]
+    with open("updates/student_feedback.json", "w") as f:
+        json.dump(feedback, f)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0053/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0053"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0063/_env_builder_impl.py

@@ -0,0 +1,109 @@
+import os
+import argparse
+import json
+import shutil
+
+def write_json(path, data):
+    with open(path, "w", encoding="utf-8") as f:
+        json.dump(data, f, indent=2)
+
+def build_turn_1():
+    os.makedirs("submissions", exist_ok=True)
+    os.makedirs("results", exist_ok=True)
+    
+    artworks = [
+        # Valid, cheap, but is a sculpture (will be banned in turn 2)
+        {"id": "art_01", "artist_name": "John Smith", "state": "AR", "medium": "Sculpture", "price": 150, "weight_lbs": 30, "tags": ["peace"]},
+        # Valid, very cheap, will be selected
+        {"id": "art_02", "artist_name": "Jane Doe", "state": "AR", "medium": "Watercolor", "price": 200, "weight_lbs": 2, "tags": ["nature"]},
+        # Invalid state
+        {"id": "art_03", "artist_name": "Bob Ross", "state": "TX", "medium": "Oil Painting", "price": 100, "weight_lbs": 5, "tags": ["quiet"]},
+        # Invalid medium
+        {"id": "art_04", "artist_name": "Alice Wonderland", "state": "AR", "medium": "Digital Art", "price": 50, "weight_lbs": 1, "tags": ["home"]},
+        # Valid, cheap, will be selected
+        {"id": "art_05", "artist_name": "Charlie Smith", "state": "AR", "medium": "Acrylic", "price": 300, "weight_lbs": 8, "tags": ["home", "quiet"]},
+        # Valid, expensive
+        {"id": "art_06", "artist_name": "Diana Prince", "state": "AR", "medium": "Oil Painting", "price": 1200, "weight_lbs": 15, "tags": ["peace"]},
+        # Valid, cheap sculpture (will be banned in turn 2 due to weight > 20)
+        {"id": "art_07", "artist_name": "Evan Miller", "state": "AR", "medium": "Sculpture", "price": 250, "weight_lbs": 25, "tags": ["nature"]},
+        # Valid, moderate price
+        {"id": "art_08", "artist_name": "Fiona Apple", "state": "AR", "medium": "Oil Painting", "price": 400, "weight_lbs": 10, "tags": ["quiet"]},
+        # Valid, cheap, but heavy painting (will be banned in turn 2)
+        {"id": "art_09", "artist_name": "George Miller", "state": "AR", "medium": "Acrylic", "price": 350, "weight_lbs": 22, "tags": ["home"]},
+        # Valid, moderate
+        {"id": "art_10", "artist_name": "Hannah Abbott", "state": "AR", "medium": "Watercolor", "price": 300, "weight_lbs": 4, "tags": ["peace"]},
+        # Valid, cheap sculpture, weight < 20 (but still banned in turn 2 because it's a sculpture)
+        {"id": "art_11", "artist_name": "Ian McKellen", "state": "AR", "medium": "Sculpture", "price": 100, "weight_lbs": 10, "tags": ["home"]}
+    ]
+    
+    # In Turn 1: 
+    # Budget 2000. Goal: max pieces.
+    # Valid pool:
+    # art_11: 100
+    # art_01: 150
+    # art_02: 200
+    # art_07: 250
+    # art_05: 300
+    # art_10: 300
+    # art_09: 350
+    # art_08: 400
+    # Total for these 8 = 2050. We can pick 7 pieces to maximize count.
+    # We drop the most expensive (art_08, 400). Total spent = 1650.
+    
+    for art in artworks:
+        write_json(f"submissions/{art['id']}.json", art)
+
+def build_turn_2():
+    # Assume previous state is kept by the framework.
+    os.makedirs("new_submissions", exist_ok=True)
+    
+    new_artworks = [
+        # Valid spiritual replacement, cheap
+        {"id": "art_12", "artist_name": "Kevin Doe", "state": "AR", "medium": "Watercolor", "price": 200, "weight_lbs": 3, "tags": ["spiritual", "quiet"]},
+        # Valid spiritual replacement, moderate
+        {"id": "art_13", "artist_name": "Laura Smith", "state": "AR", "medium": "Oil Painting", "price": 400, "weight_lbs": 12, "tags": ["faith", "peace"]},
+        # Invalid: spiritual but it's a sculpture
+        {"id": "art_14", "artist_name": "Mike Johnson", "state": "AR", "medium": "Sculpture", "price": 150, "weight_lbs": 5, "tags": ["spiritual", "home"]},
+        # Invalid: spiritual but heavy
+        {"id": "art_15", "artist_name": "Nancy Drew", "state": "AR", "medium": "Acrylic", "price": 300, "weight_lbs": 25, "tags": ["faith", "nature"]},
+        # Valid, but too expensive to include while maximizing count
+        {"id": "art_16", "artist_name": "Oscar Wilde", "state": "AR", "medium": "Watercolor", "price": 1500, "weight_lbs": 5, "tags": ["spiritual", "peace"]}
+    ]
+    
+    for art in new_artworks:
+        write_json(f"new_submissions/{art['id']}.json", art)
+
+def build_turn_3():
+    # The final list going into turn 3 should be:
+    # Non-banned from turn 1: art_02(200), art_05(300), art_10(300), art_08(400) -> total 1200
+    # Spiritual additions from turn 2: art_12(200), art_13(400) -> total 600
+    # Grand total: 1800. 6 pieces total.
+    # Note names:
+    # art_02: Jane Doe
+    # art_05: Charlie Smith
+    # art_10: Hannah Abbott
+    # art_08: Fiona Apple
+    # art_12: Kevin Doe
+    # art_13: Laura Smith
+    # Doe family: art_02, art_12
+    # Smith family: art_05, art_13
+    
+    venue_layout = {
+        "Wall_North": {"max_pieces": 3, "max_weight_lbs": 20},
+        "Wall_East": {"max_pieces": 2, "max_weight_lbs": 15},
+        "Wall_West": {"max_pieces": 2, "max_weight_lbs": 25}
+    }
+    
+    write_json("venue_layout.json", venue_layout)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0063/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0063"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0065/_env_builder_impl.py

@@ -0,0 +1,70 @@
+import os
+import argparse
+import json
+import random
+
+def build_turn_1():
+    # 建立初始行政环境
+    os.makedirs("incident_logs", exist_ok=True)
+    os.makedirs("personnel", exist_ok=True)
+    os.makedirs("inventory", exist_ok=True)
+    
+    # 警员名单与状态，包含极其复杂的排班偏好与工会规则冲突
+    officers = [
+        {"id": "PO-101", "name": "Garcia", "rank": "Senior", "shift_preference": "Night", "hourly_rate": 65, "extra_skills": ["Spanish-Fluent", "Crisis-Negotiation"]},
+        {"id": "PO-102", "name": "O'Malley", "rank": "Junior", "shift_preference": "Day", "hourly_rate": 45, "extra_skills": []},
+        {"id": "PO-103", "name": "Chen", "rank": "Senior", "shift_preference": "Day", "hourly_rate": 68, "extra_skills": ["Forensics"]},
+        {"id": "PO-104", "name": "Smith", "rank": "Mid", "shift_preference": "Night", "hourly_rate": 55, "extra_skills": ["K9-Unit"]},
+    ]
+    with open("personnel/active_roster.json", "w") as f:
+        json.dump(officers, f)
+
+    # 包含脏数据的巡逻车日志
+    vehicle_logs = [
+        "DATE,UNIT_ID,MILEAGE_START,MILEAGE_END,FUEL_LEVEL,ISSUE_REPORTED",
+        "2023-10-01,V-99,120500,120650,45%,None",
+        "2023-10-01,V-42,45000,45120,80%,Brakes squeaking",
+        "2023-10-02,V-99,120650,ERROR_CODE_X,30%,Check Engine", # 脏数据
+        "2023-10-02,V-42,45120,45300,10%,Flat tire",
+        "2023-10-03,V-99,120650,120800,20%,Severe transmission slip" # 里程回退错误
+    ]
+    with open("inventory/patrol_vehicles.csv", "w") as f:
+        f.write("\n".join(vehicle_logs))
+
+    # 复杂规则说明：州政府年度预算剩余与加班限制（文字描述）
+    regulations = """
+    DEPARTMENTAL DIRECTIVE 2023-04:
+    1. Total remaining overtime budget for Q4: $12,500. 
+    2. Overtime is calculated as 1.5x hourly rate after 8 hours in a 24h window.
+    3. Vehicles with mileage over 120,000 must undergo Level 2 Inspection before any pursuit-grade missions.
+    4. Junior officers (Rank: Junior) must never be paired together on Night shifts.
+    """
+    with open("internal_policy.txt", "w") as f:
+        f.write(regulations)
+
+def build_turn_2():
+    # 增加新的冲突：紧急调配与预算变动
+    os.makedirs("emergency_orders", exist_ok=True)
+    new_task = {
+        "event": "Public Demonstration Support",
+        "date": "2023-10-15",
+        "required_officers": 3,
+        "special_requirement": "Bilingual-Spanish Preferred",
+        "est_duration_hours": 12
+    }
+    with open("emergency_orders/demo_task.json", "w") as f:
+        json.dump(new_task, f)
+
+def build_turn_3():
+    # 第三轮：设备彻底报废导致的连带责任分析
+    # 此时 Agent 应该已经记录了 V-99 的问题
+    with open("incident_logs/incident_report_1016.txt", "w") as f:
+        f.write("OFFICER: O'Malley. VEHICLE: V-99. STATUS: Total Loss. ACCIDENT: Transmission failure during routine patrol. Budget for new vehicle replacement needed.")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    if args.turn == 1: build_turn_1()
+    elif args.turn == 2: build_turn_2()
+    elif args.turn == 3: build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0065/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0065"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0075/_env_builder_impl.py

@@ -0,0 +1,93 @@
+import os
+import argparse
+import csv
+import json
+
+def build_turn_1():
+    # 创建目录结构
+    os.makedirs("orders", exist_ok=True)
+    os.makedirs("chemicals", exist_ok=True)
+    os.makedirs("staff", exist_ok=True)
+    os.makedirs("factory_management", exist_ok=True)
+    os.makedirs("memos", exist_ok=True)
+
+    # 1. 常规订单数据
+    orders = [
+        {"order_id": "ORD_001", "product": "Dining_Table", "finish_required": "Clear_Coat", "est_hours": 5},
+        {"order_id": "ORD_002", "product": "Oak_Chairs", "finish_required": "Stain", "est_hours": 8},
+        {"order_id": "ORD_003", "product": "Bookshelf", "finish_required": "Paint", "est_hours": 4}
+    ]
+    with open("orders/regular.csv", "w", newline='') as f:
+        writer = csv.DictWriter(f, fieldnames=["order_id", "product", "finish_required", "est_hours"])
+        writer.writeheader()
+        writer.writerows(orders)
+
+    # 2. 化学品库存 (VOC上限450，由prompt给出)
+    # 陷阱：Chem_D 满足Stain需求，且VOC 350合法，但含有 Toluene。第一轮必须选它，第二轮它会被ban。
+    # 陷阱：Chem_C 是Stain，但是 VOC 480，第一轮就不合法。
+    chemicals = [
+        {"chem_id": "Chem_A", "type": "Clear_Coat", "voc_g_L": 400, "contains_toluene": True},
+        {"chem_id": "Chem_B", "type": "Clear_Coat", "voc_g_L": 300, "contains_toluene": False},
+        {"chem_id": "Chem_C", "type": "Stain", "voc_g_L": 480, "contains_toluene": False},
+        {"chem_id": "Chem_D", "type": "Stain", "voc_g_L": 350, "contains_toluene": True},
+        {"chem_id": "Chem_E", "type": "Paint", "voc_g_L": 410, "contains_toluene": False}
+    ]
+    with open("chemicals/inventory.json", "w") as f:
+        json.dump(chemicals, f, indent=4)
+
+    # 3. 员工状态 (上限40小时，由prompt给出)
+    # 第一轮消耗计算：
+    # ORD_001 (5h) -> 只能分给 Carlos (32+5=37) 或 Luis (30+5=35)。Maria不够 (36+5=41)。
+    # ORD_002 (8h) -> 只能分给 Carlos (32+8=40) 或 Luis (30+8=38)。如果分给Carlos，他刚好满40。
+    # ORD_003 (4h) -> Maria (36+4=40) 刚好可以。
+    staff = [
+        {"name": "Carlos", "role": "Mixer", "hours_worked_so_far": 32},
+        {"name": "Maria", "role": "Polisher", "hours_worked_so_far": 36},
+        {"name": "Luis", "role": "General", "hours_worked_so_far": 30}
+    ]
+    with open("staff/team.csv", "w", newline='') as f:
+        writer = csv.DictWriter(f, fieldnames=["name", "role", "hours_worked_so_far"])
+        writer.writeheader()
+        writer.writerows(staff)
+
+
+def build_turn_2():
+    # 此阶段运行在 turn_2 的工作区，turn_1产生的文件(备忘录、计划)已被框架复制过来
+    os.makedirs("memos", exist_ok=True)
+    os.makedirs("orders", exist_ok=True)
+    os.makedirs("chemicals", exist_ok=True)
+
+    # 1. 突发信件 (宣布 Toluene 违规)
+    with open("memos/boss_note.txt", "w") as f:
+        f.write("ROSA! The OSHA inspector is physically in the building!\n")
+        f.write("Immediate directive: ANY chemical containing 'Toluene' is BANNED from use on the floor starting right now. DO NOT USE IT.\n")
+        f.write("If we have orders assigned to Toluene chemicals, swap them out immediately before they see the logs!\n")
+
+    # 2. 紧急订单 (需要依靠剩余可用工时)
+    # RUSH_001 需要 Stain, 3小时。
+    rush_orders = [
+        {"order_id": "RUSH_001", "product": "Coffee_Table", "finish_required": "Stain", "est_hours": 3}
+    ]
+    with open("orders/rush.csv", "w", newline='') as f:
+        writer = csv.DictWriter(f, fieldnames=["order_id", "product", "finish_required", "est_hours"])
+        writer.writeheader()
+        writer.writerows(rush_orders)
+
+    # 3. 新到的化学品 (拯救因为 Toluene 被ban而无法完成的 Stain 订单)
+    # Chem_F: Stain, VOC 420 (合法), 无 Toluene (合法)。
+    new_chemicals = [
+        {"chem_id": "Chem_F", "type": "Stain", "voc_g_L": 420, "contains_toluene": False}
+    ]
+    with open("chemicals/new_arrival.json", "w") as f:
+        json.dump(new_chemicals, f, indent=4)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0075/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0075"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0077/_env_builder_impl.py

@@ -0,0 +1,106 @@
+import os
+import argparse
+import json
+import csv
+import xml.etree.ElementTree as ET
+
+def build_turn_1():
+    # Create directories
+    os.makedirs("ward_a_patients", exist_ok=True)
+    os.makedirs("medical_refs", exist_ok=True)
+    os.makedirs("garden_catalog", exist_ok=True)
+
+    # 1. Ward A Patients (JSON)
+    patients_a = [
+        {"id": "A1", "name": "Mr. Smith", "age": 78, "medications": ["Lisinopril", "Aspirin"]},
+        {"id": "A2", "name": "Mrs. Higgins", "age": 82, "medications": ["Metformin", "Vitamin D"]},
+        {"id": "A3", "name": "Mr. Davis", "age": 69, "medications": ["Atorvastatin", "Omeprazole"]}
+    ]
+    for p in patients_a:
+        with open(f"ward_a_patients/patient_{p['id']}.json", "w") as f:
+            json.dump(p, f, indent=4)
+
+    # 2. Herb Contraindications (CSV) - Contains traps for Turn 2
+    contraindications = [
+        ["Herb", "Medication", "Risk_Level"],
+        ["Licorice Root", "Lisinopril", "High"],
+        ["Ginseng", "Metformin", "Severe"],
+        ["St. John's Wort", "Atorvastatin", "Moderate"],
+        ["Spinach", "Warfarin", "Severe"],
+        ["Kale", "Warfarin", "Severe"],
+        ["Chamomile", "Ibuprofen", "Moderate"],
+        ["Rosemary", "Aspirin", "Low"],
+        ["Thyme", "Penicillin", "Low"]
+    ]
+    with open("medical_refs/herb_contraindications.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerows(contraindications)
+
+    # 3. Spring Seeds Catalog (XML)
+    # Target lowest safe cost: Spinach (8) + Basil (12) + Mint (15) = 35.
+    # Trap: Spinach is extremely cheap and perfectly safe in Turn 1, but deadly in Turn 2.
+    root = ET.Element("Catalog", season="Spring")
+    seeds = [
+        {"name": "Mint", "cost": "15", "difficulty": "2"},
+        {"name": "Basil", "cost": "12", "difficulty": "1"},
+        {"name": "Lavender", "cost": "22", "difficulty": "3"},
+        {"name": "Spinach", "cost": "8", "difficulty": "1"},
+        {"name": "Licorice Root", "cost": "10", "difficulty": "2"},
+        {"name": "Rosemary", "cost": "18", "difficulty": "4"}
+    ]
+    for s in seeds:
+        item = ET.SubElement(root, "Seed")
+        ET.SubElement(item, "Name").text = s["name"]
+        ET.SubElement(item, "Cost").text = s["cost"]
+        ET.SubElement(item, "Difficulty").text = s["difficulty"]
+    
+    tree = ET.ElementTree(root)
+    tree.write("garden_catalog/spring_seeds.xml")
+
+def build_turn_2():
+    # Directories might already exist if framework copies workspace, but we ensure existence for new ones
+    os.makedirs("ward_b_patients", exist_ok=True)
+    os.makedirs("updates", exist_ok=True)
+    os.makedirs("garden_catalog", exist_ok=True)
+
+    # 1. Ward B Patients (JSON)
+    patients_b = [
+        {"id": "B1", "name": "Mrs. Clark", "age": 71, "medications": ["Ibuprofen", "Calcium"]},
+        {"id": "B2", "name": "Mr. Lewis", "age": 88, "medications": ["Amlodipine"]}
+    ]
+    for p in patients_b:
+        with open(f"ward_b_patients/patient_{p['id']}.json", "w") as f:
+            json.dump(p, f, indent=4)
+
+    # 2. Updates (TXT)
+    with open("updates/ward_a_changes.txt", "w") as f:
+        f.write("URGENT MEDICAL UPDATE:\n")
+        f.write("Mrs. Higgins (Patient A2) experienced a TIA last night.\n")
+        f.write("Orders from Dr. Evans:\n")
+        f.write("- Discontinue Metformin immediately.\n")
+        f.write("- Begin Warfarin 5mg daily to manage clotting risk.\n")
+        f.write("Please update all charts and dietary plans accordingly.\n")
+
+    # 3. Summer Seeds Catalog (JSON)
+    # Agent needs 2 more seeds to reach 4 (since Spinach is discarded, leaving Mint and Basil).
+    # Safe options for Ward A & B combined: Thyme, Oregano, Parsley.
+    # Lowest cost for 2 safe summer seeds: Parsley (9) + Thyme (10) = 19.
+    summer_seeds = [
+        {"name": "Thyme", "cost": 10, "difficulty": 2},
+        {"name": "Oregano", "cost": 14, "difficulty": 2},
+        {"name": "Parsley", "cost": 9, "difficulty": 1},
+        {"name": "Kale", "cost": 7, "difficulty": 1},      # Unsafe (Warfarin)
+        {"name": "Chamomile", "cost": 13, "difficulty": 2} # Unsafe (Ibuprofen)
+    ]
+    with open("garden_catalog/summer_seeds.json", "w") as f:
+        json.dump(summer_seeds, f, indent=4)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0077/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0077"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0079/_env_builder_impl.py

@@ -0,0 +1,66 @@
+import os
+import argparse
+import json
+import random
+
+def build_turn_1():
+    # 核心代码库模拟：包含多个微服务源码
+    services = ["auth-module", "payment-gateway", "data-aggregator"]
+    os.makedirs("src", exist_ok=True)
+    
+    for svc in services:
+        svc_path = os.path.join("src", svc)
+        os.makedirs(svc_path, exist_ok=True)
+        # 写入混有注释、复杂逻辑和隐患的代码
+        with open(os.path.join(svc_path, "core.py"), "w") as f:
+            f.write(f'# Service: {svc}\n')
+            f.write('import hashlib\n\n')
+            f.write('def process_request(data):\n')
+            if svc == "auth-module":
+                f.write('    # Legacy: Using MD5 for quick hash, must be replaced later\n')
+                f.write('    token = hashlib.md5(data.encode()).hexdigest()\n')
+                f.write('    return {"status": "ok", "token": token}\n')
+            elif svc == "payment-gateway":
+                f.write('    # Dependency: requires auth-module token validation\n')
+                f.write('    if len(data.get("token", "")) < 32:\n')
+                f.write('        return {"error": "invalid"}\n')
+                f.write('    return {"amount": 100, "currency": "USD"}\n')
+            else:
+                f.write('    return {"data": "processed"}\n')
+
+    # 遗留资产清单（干扰项）
+    with open("legacy_inventory.json", "w") as f:
+        json.dump({
+            "obsolete_servers": ["SRV-001", "SRV-099"],
+            "tech_stack": ["Python 2.7", "Python 3.10"],
+            "critical_dependencies": ["auth-module", "payment-gateway"]
+        }, f, indent=4)
+
+def build_turn_2():
+    # 模拟外部合规文件更新
+    os.makedirs("compliance", exist_ok=True)
+    with open("compliance/security_policy_v2.txt", "w") as f:
+        f.write("SECURITY POLICY VERSION 2.0\n")
+        f.write("---------------------------\n")
+        f.write("1. All cryptographic hashing must use SHA-256 or higher.\n")
+        f.write("2. MD5 is strictly prohibited for sensitive data tokens.\n")
+        f.write("3. Deprecated modules must be flagged in a separate migration-plan.")
+
+def build_turn_3():
+    # 注入突发错误日志，指向受影响的服务
+    os.makedirs("logs", exist_ok=True)
+    with open("logs/runtime_error.log", "w") as f:
+        f.write("2023-10-27 10:15:02 ERROR: Validation failed in payment-gateway. Reason: Token format mismatch.\n")
+        f.write("2023-10-27 10:16:45 WARN: auth-module response time exceeding 200ms.\n")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0079/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0079"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0082/_env_builder_impl.py

@@ -0,0 +1,112 @@
+import os
+import json
+import csv
+import argparse
+
+def build_turn_1():
+    os.makedirs("ward_data", exist_ok=True)
+    os.makedirs("inventory", exist_ok=True)
+    os.makedirs("reports", exist_ok=True)
+
+    patients = [
+        ["id", "name", "age", "diet_code", "allergies", "mobility_status"],
+        ["P01", "John Smith", "65", "D1", "None", "Walking"],
+        ["P02", "Alice Johnson", "72", "D2", "dairy", "Bedbound"],
+        ["P03", "Robert Davis", "58", "D3", "nuts", "Wheelchair"],
+        ["P04", "Emma Brown", "81", "D1", "None", "Walking"],
+        ["P05", "Michael Miller", "45", "D2", "None", "Walking"]
+    ]
+    with open("ward_data/patients.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerows(patients)
+
+    codes = (
+        "Diet Codes Explanation:\n"
+        "D1: Standard Diabetic Diet\n"
+        "D2: Renal Diet (Requires Renal_Supplement)\n"
+        "D3: Liquid Diet (Requires Liquid_Supplement)\n"
+    )
+    with open("ward_data/codes.txt", "w") as f:
+        f.write(codes)
+
+    catalog = [
+        {"item_name": "Renal_Supp_A", "type": "Renal_Supplement", "contains_allergens": ["dairy", "soy"]},
+        {"item_name": "Renal_Supp_B", "type": "Renal_Supplement", "contains_allergens": ["gluten"]},
+        {"item_name": "Liquid_Supp_X", "type": "Liquid_Supplement", "contains_allergens": ["nuts", "dairy"]},
+        {"item_name": "Liquid_Supp_Y", "type": "Liquid_Supplement", "contains_allergens": []}
+    ]
+    with open("inventory/catalog.json", "w") as f:
+        json.dump(catalog, f, indent=4)
+
+    stock = {
+        "Renal_Supp_A": 20,
+        "Renal_Supp_B": 5,
+        "Liquid_Supp_X": 10,
+        "Liquid_Supp_Y": 2
+    }
+    with open("inventory/stock.json", "w") as f:
+        json.dump(stock, f, indent=4)
+
+def build_turn_2():
+    os.makedirs("updates", exist_ok=True)
+    
+    # 陷阱: 新来的海地病人 P06, 对 soy 过敏。需要 Renal。
+    new_patients = [
+        ["id", "name", "age", "diet_code", "allergies", "mobility_status"],
+        ["P06", "Jean Baptiste", "68", "D2", "soy", "Bedbound"],
+        ["P07", "Sarah Connor", "40", "D3", "None", "Walking"]
+    ]
+    with open("updates/new_patients.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerows(new_patients)
+
+    # 陷阱: Renal_Supp_B 最便宜，但不适合 P06 (如果是买别的)。
+    # 稍微调整：旧 catalog 里 Renal_Supp_B 含 gluten，适合 P06 和 P02。
+    # 现在给新品牌定价。
+    new_prices = [
+        {"item_name": "Renal_Supp_B", "price": 4.0, "allergen_traces": ["gluten"]},
+        {"item_name": "Renal_Supp_C", "price": 2.5, "allergen_traces": ["soy"]}, # 最便宜，但 P06 过敏
+        {"item_name": "Renal_Supp_D", "price": 6.0, "allergen_traces": []},
+        {"item_name": "Liquid_Supp_Y", "price": 5.0, "allergen_traces": []},
+        {"item_name": "Liquid_Supp_Z", "price": 3.0, "allergen_traces": ["nuts"]} # P03 对 nuts 过敏
+    ]
+    with open("updates/new_prices.json", "w") as f:
+        json.dump(new_prices, f, indent=4)
+
+def build_turn_3():
+    os.makedirs("staff", exist_ok=True)
+    
+    trainings = [
+        ["staff_id", "name", "certifications"],
+        ["S01", "Nancy Wheeler", "Dietary_Safety;Basic_CPR"],
+        ["S02", "Steve Harrington", "Dietary_Safety;Allergy_Care;Basic_CPR"],
+        ["S03", "Robin Buckley", "Allergy_Care"],
+        ["S04", "Dustin Henderson", "Dietary_Safety;Allergy_Care"]
+    ]
+    with open("staff/trainings.csv", "w", newline="") as f:
+        writer = csv.writer(f)
+        writer.writerows(trainings)
+
+    # 陷阱: S01 只照顾了 P02 (有过敏)，缺 Allergy_Care。 S03 照顾了 P06 (新来的老乡，有过敏)，缺 Dietary_Safety。
+    schedule = {
+        "Monday_Shift": [
+            {"staff_id": "S01", "assigned_patients": ["P01", "P02", "P04"]},
+            {"staff_id": "S02", "assigned_patients": ["P03"]},
+            {"staff_id": "S03", "assigned_patients": ["P05", "P06"]},
+            {"staff_id": "S04", "assigned_patients": ["P07"]}
+        ]
+    }
+    with open("staff/schedule.json", "w") as f:
+        json.dump(schedule, f, indent=4)
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--turn", type=int, required=True)
+    args = parser.parse_args()
+    
+    if args.turn == 1:
+        build_turn_1()
+    elif args.turn == 2:
+        build_turn_2()
+    elif args.turn == 3:
+        build_turn_3()

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0082/env_builder.py

@@ -0,0 +1,17 @@
+from __future__ import annotations
+
+import os
+import runpy
+from pathlib import Path
+
+
+def main() -> None:
+    repo_root = Path(__file__).resolve().parents[2]
+    asset_dir = repo_root / "assets" / "data_round_01_aligned_mix_800_0082"
+    asset_dir.mkdir(parents=True, exist_ok=True)
+    os.chdir(asset_dir)
+    runpy.run_path(str(Path(__file__).with_name("_env_builder_impl.py")), run_name="__main__")
+
+
+if __name__ == "__main__":
+    main()