Add files using upload-large-folder tool

.gitignore

@@ -0,0 +1,6 @@
+.cache/
+__pycache__/
+*.pyc
+*.pyo
+round_01_aligned_mix_800/assets/
+persona_aligned_mix_200/assets/

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0025.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0025
+name: social_service_resource_orchestration
+description: 测试 Agent 在处理复杂医疗社会工作资源调配时的状态流转能力。涉及多维度约束计算（预算、合规性、家庭匹配度）、动态数据注入后的历史规则追溯，以及多文件环境下的状态保留。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0025_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0025_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0025_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0025
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0025/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0025/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0025/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0025_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0025_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0025_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0046.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0046
+name: school_community_festival_coordinator
+description: 模拟一位小学助教协调社区文化节。任务涉及处理多源志愿者工时核对、预算动态超支预判、以及基于历史约束的复杂供应商筛选。测试 Agent 的跨会话规则记忆、多文件逻辑关联计算及应对突发冲突的能力。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0046_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0046_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0046_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0046
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0046/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0046/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0046/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0046_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0046_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0046_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0047.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0047
+name: anime_culinary_philosophy_retreat
+description: 测试 Agent 在多轮复杂约束下的状态持久化、多文件交叉推理与动态调整能力。涉及多条件的预算计算、时间规划以及由于突发事件（文件遗失、设备故障、新数据注入）导致的重规划。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0047_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0047_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0047_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0047
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0047/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0047/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0047/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0047_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0047_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0047_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0088.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0088
+name: wearable_tech_logistics_optimization
+description: 针对智能可穿戴时装发布会的多轮物流优化任务，测试 Agent 在复杂业务逻辑下的状态流转、数据冲突处理及长期规则记忆能力。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0088_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0088_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0088_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0088
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0088/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0088/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0088/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0088_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0088_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0088_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0093.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0093
+name: logistics_route_optimization_and_compliance
+description: 针对一名危地马拉裔资深物流调度员的日常工作，模拟多轮复杂的配送路径规划、成本核算及突发合规性审查。测试 Agent 在多变规则下的逻辑推理、文件状态保留及增量数据处理能力。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0093_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0093_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0093_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0093
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0093/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0093/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0093/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0093_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0093_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0093_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0100.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0100
+name: poetry_club_logistics_management
+description: 协助一名极度严谨且内向的高一学生管理文学社的年度诗集出版与财务对账，涉及跨文件的复杂逻辑校验、历史规则继承与增量冲突处理。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0100_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0100_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0100_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0100
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0100/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0100/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0100/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0100_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0100_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0100_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0119.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0119
+name: vietnamese_machinist_chaos
+description: 测试 Agent 在复杂状态流转中的资源重分配、依赖记忆消除冲突及历史溯源统计能力
+prompts:
+- prompts/data_round_01_aligned_mix_800_0119_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0119_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0119_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0119
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0119/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0119/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0119/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0119_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0119_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0119_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0139.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0139
+name: martha_multicultural_festival
+description: 模拟政府行政助理在复杂多变环境下的长程任务处理。测试Agent能否在多轮会话中通过物理文件记录、更新状态，并在突发需求（预算缩减、场地关闭）下严格遵守未在当前Prompt重申的历史业务红线（如卫生评级、特定饮食需求限制），同时进行多维度的复杂数据筛选与统筹规划。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0139_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0139_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0139_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0139
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0139/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0139/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0139/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0139_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0139_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0139_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0146.yaml

@@ -0,0 +1,27 @@
+id: data_round_01_aligned_mix_800_0146
+name: wholesale_inventory_dynamic_compliance
+description: 评估 Agent 在机械批发管理中的多轮库存审计、合规性流转及突发环境变更的处理能力。Agent 需要在第一轮建立复杂的库存规则与分配逻辑，并在第二轮处理规则冲突与供应链突发数据，且必须依赖第一轮自行建立的记录文件。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0146_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0146_turn_2.md
+environment:
+  asset: data_round_01_aligned_mix_800_0146
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0146/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0146/turn_2
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0146_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0146_turn_2.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0187.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0187
+name: construction_logistics_and_compliance
+description: 评估 Agent 在高压工程环境下的多轮状态流转能力。涉及处理复杂的施工进度表、预算超支预警、以及在不重申历史规则的情况下，根据前序轮次设定的合规标准处理新突发的供应冲突。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0187_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0187_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0187_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0187
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0187/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0187/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0187/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0187_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0187_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0187_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0195.yaml

@@ -0,0 +1,32 @@
+id: data_round_01_aligned_mix_800_0195
+name: renewable_grid_expansion_management
+description: 测试Agent在电力系统扩容项目中的多轮状态流转能力。涉及复杂的工程选址、合规性校验及应对突发政策变更，要求Agent自行记录选址准则并在后续轮次中自主维持逻辑一致性。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0195_turn_1.md
+- prompts/data_round_01_aligned_mix_800_0195_turn_2.md
+- prompts/data_round_01_aligned_mix_800_0195_turn_3.md
+environment:
+  asset: data_round_01_aligned_mix_800_0195
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+    turn_1:
+    - data_round_01_aligned_mix_800_0195/turn_1
+    turn_2:
+    - data_round_01_aligned_mix_800_0195/turn_2
+    turn_3:
+    - data_round_01_aligned_mix_800_0195/turn_3
+sessions:
+- turn: 1
+  prompt: prompts/data_round_01_aligned_mix_800_0195_turn_1.md
+- turn: 2
+  prompt: prompts/data_round_01_aligned_mix_800_0195_turn_2.md
+- turn: 3
+  prompt: prompts/data_round_01_aligned_mix_800_0195_turn_3.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0206.yaml

@@ -0,0 +1,22 @@
+id: data_round_01_aligned_mix_800_0206
+name: data_round_01_aligned_mix_800_0206
+description: Act as a stressed, highly critical preschool teacher asking the agent
+  to parse messy fundraiser files for specific donation statistics and VIP parent
+  names. Requires classification and external ledger lookup.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0206.md
+environment:
+  asset: data_round_01_aligned_mix_800_0206
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0206-educational-content-classifier-skill
+  - data-round-01-aligned-mix-800-0206-fast-donations-summarizer
+  - data-round-01-aligned-mix-800-0206-school-parent-ledger-checker
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+prompt: prompts/data_round_01_aligned_mix_800_0206.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0229.yaml

@@ -0,0 +1,22 @@
+id: data_round_01_aligned_mix_800_0229
+name: library_digitization_integrity_check
+description: A library assistant needs help reconciling a messy physical archive digitization
+  project, involving metadata validation, duplicate detection, and cost estimation
+  using legacy APIs.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0229.md
+environment:
+  asset: data_round_01_aligned_mix_800_0229
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0229-archival-decoder-skill
+  - data-round-01-aligned-mix-800-0229-v1-restoration-cost-estimator-skill
+  - data-round-01-aligned-mix-800-0229-v2-restoration-cost-estimator-skill
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+category: Data Processing

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0247.yaml

@@ -0,0 +1,25 @@
+id: data_round_01_aligned_mix_800_0247
+name: data_round_01_aligned_mix_800_0247
+description: 一个日本裔美国失业男士需要处理一份混乱的社区志愿者数据库，解析加密日志，并使用外部API查明恶意签到的动漫假名。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0247.md
+environment:
+  asset: data_round_01_aligned_mix_800_0247
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0247-akihabara-deep-search
+  - data-round-01-aligned-mix-800-0247-myanimelist-api-client
+  - data-round-01-aligned-mix-800-0247-st-catharina-dat-parser
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+category: Data processing
+tags:
+- python
+- data-cleaning
+- report-generation
+- tool-usage

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0256.yaml

@@ -0,0 +1,22 @@
+id: data_round_01_aligned_mix_800_0256
+name: data_round_01_aligned_mix_800_0256
+prompts:
+- prompts/data_round_01_aligned_mix_800_0256.md
+environment:
+  asset: data_round_01_aligned_mix_800_0256
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0256-heritage-catalog-search
+  - data-round-01-aligned-mix-800-0256-vintage-db-query
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: '1.0'
+prompt: prompts/data_round_01_aligned_mix_800_0256.md
+dependencies:
+- openai
+- httpx

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0276.yaml

@@ -0,0 +1,30 @@
+id: data_round_01_aligned_mix_800_0276
+name: data_round_01_aligned_mix_800_0276
+description: The agent must process messy expense records (including a proprietary
+  encrypted .osnas file), query a Tax Oracle API to dynamically determine deductibility
+  rules, calculate totals, and flag habitual offenders.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0276.md
+environment:
+  asset: data_round_01_aligned_mix_800_0276
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0276-legacy-tax-oracle
+  - data-round-01-aligned-mix-800-0276-osnas-decryptor
+  - data-round-01-aligned-mix-800-0276-q3-tax-oracle
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: '1.0'
+type: evaluation
+task_id: data_round_01_aligned_mix_800_0276
+title: Tax Compliance Automation for a Gadget-Obsessed Manager
+assets:
+- data_round_01_aligned_mix_800_0276
+prompt_file: prompts/data_round_01_aligned_mix_800_0276.md
+verify_rules_file: tasks/data_round_01_aligned_mix_800_0276/verify_rules.py
+verify_prompt_file: tasks/data_round_01_aligned_mix_800_0276/verify_prompt.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0296.yaml

@@ -0,0 +1,19 @@
+id: data_round_01_aligned_mix_800_0296
+name: data_round_01_aligned_mix_800_0296
+prompts:
+- prompts/data_round_01_aligned_mix_800_0296.md
+environment:
+  asset: data_round_01_aligned_mix_800_0296
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0296-legacy-inventory-check
+  - data-round-01-aligned-mix-800-0296-query-wms-database
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: '1.0'
+prompt: prompts/data_round_01_aligned_mix_800_0296.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0297.yaml

@@ -0,0 +1,22 @@
+id: data_round_01_aligned_mix_800_0297
+name: data_round_01_aligned_mix_800_0297
+description: Analyze student botany photo logs using vision skills and verify native
+  plant growth against an external species database.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0297.md
+environment:
+  asset: data_round_01_aligned_mix_800_0297
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0297-botanical-image-analyzer-skill
+  - data-round-01-aligned-mix-800-0297-global-botany-lookup-skill
+  - data-round-01-aligned-mix-800-0297-open-weed-wiki-skill
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+prompt: prompts/data_round_01_aligned_mix_800_0297.md
+timeout: 450

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0301.yaml

@@ -0,0 +1,21 @@
+id: data_round_01_aligned_mix_800_0301
+name: data_round_01_aligned_mix_800_0301
+prompts:
+- prompts/data_round_01_aligned_mix_800_0301.md
+environment:
+  asset: data_round_01_aligned_mix_800_0301
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0301-global-search-engine-skill
+  - data-round-01-aligned-mix-800-0301-mariachi-finance-validator-skill
+  - data-round-01-aligned-mix-800-0301-pdf-parser-skill
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+schema_version: v1
+prompt: prompts/data_round_01_aligned_mix_800_0301.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0319.yaml

@@ -0,0 +1,22 @@
+id: data_round_01_aligned_mix_800_0319
+name: data_round_01_aligned_mix_800_0319
+description: Assist a stressed, disorganized Vietnamese-American machinist in compiling
+  a machine maintenance report by decoding proprietary diagnostic files and querying
+  ERP tools.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0319.md
+environment:
+  asset: data_round_01_aligned_mix_800_0319
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0319-cloud-erp-query
+  - data-round-01-aligned-mix-800-0319-cnc-diag-decoder
+  - data-round-01-aligned-mix-800-0319-legacy-erp-query
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+prompt: prompts/data_round_01_aligned_mix_800_0319.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0324.yaml

@@ -0,0 +1,20 @@
+id: data_round_01_aligned_mix_800_0324
+name: data_round_01_aligned_mix_800_0324
+prompts:
+- prompts/data_round_01_aligned_mix_800_0324.md
+environment:
+  asset: data_round_01_aligned_mix_800_0324
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0324-bing-search-skill
+  - data-round-01-aligned-mix-800-0324-heritage-receipt-ocr-skill
+  - data-round-01-aligned-mix-800-0324-member-identity-validator-skill
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+task_id: data_round_01_aligned_mix_800_0324
+prompt: prompts/data_round_01_aligned_mix_800_0324.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0326.yaml

@@ -0,0 +1,28 @@
+id: data_round_01_aligned_mix_800_0326
+name: data_round_01_aligned_mix_800_0326
+description: Analyze messy community center health screening logs, decode encrypted
+  medical data, and verify volunteer eligibility via external APIs.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0326.md
+environment:
+  asset: data_round_01_aligned_mix_800_0326
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0326-bp-decrypter
+  - data-round-01-aligned-mix-800-0326-legacy-clinic-auth
+  - data-round-01-aligned-mix-800-0326-national-volunteer-db
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+category: Data Analysis
+tags:
+- data_cleaning
+- authorization_verification
+- medical_logs
+- tool_use
+- decryption
+prompt: prompts/data_round_01_aligned_mix_800_0326.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0328.yaml

@@ -0,0 +1,20 @@
+id: data_round_01_aligned_mix_800_0328
+name: data_round_01_aligned_mix_800_0328
+prompts:
+- prompts/data_round_01_aligned_mix_800_0328.md
+environment:
+  asset: data_round_01_aligned_mix_800_0328
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0328-encrypted-demographic-parser-skill
+  - data-round-01-aligned-mix-800-0328-legacy-risk-lookup-service
+  - data-round-01-aligned-mix-800-0328-risk-actuary-quantifier-skill
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+schema_version: '1.0'

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0368.yaml

@@ -0,0 +1,21 @@
+id: data_round_01_aligned_mix_800_0368
+name: data_round_01_aligned_mix_800_0368
+description: Extract and decode specific customer feedback regarding diversity and
+  accessibility from a messy retail POS data dump, requiring custom binary decoding
+  and CRM API lookups.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0368.md
+environment:
+  asset: data_round_01_aligned_mix_800_0368
+skills:
+  available:
+  - data-round-01-aligned-mix-800-0368-cloud-crm-lookup-skill
+  - data-round-01-aligned-mix-800-0368-legacy-crm-lookup-skill
+  - data-round-01-aligned-mix-800-0368-pos-binary-decoder-skill
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0419.yaml

@@ -0,0 +1,18 @@
+id: data_round_01_aligned_mix_800_0419
+name: data_round_01_aligned_mix_800_0419
+prompts:
+- prompts/data_round_01_aligned_mix_800_0419.md
+environment:
+  asset: data_round_01_aligned_mix_800_0419
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: '1.0'
+task_id: data_round_01_aligned_mix_800_0419
+prompt: prompts/data_round_01_aligned_mix_800_0419.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0430.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0430
+name: data_round_01_aligned_mix_800_0430_student_performance_wasteland
+description: A teacher needs help reconciling severely corrupted and fragmented student performance data scattered across a chaotic file system to generate a final report.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0430.md
+environment:
+  asset: data_round_01_aligned_mix_800_0430
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+prompt: prompts/data_round_01_aligned_mix_800_0430.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0436.yaml

@@ -0,0 +1,19 @@
+id: data_round_01_aligned_mix_800_0436
+name: data_round_01_aligned_mix_800_0436
+description: Reconstruct fundraiser financials from a fragmented, noise-heavy environment by filtering against a hidden whitelist and multi-source pricing logs.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0436.md
+environment:
+  asset: data_round_01_aligned_mix_800_0436
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+schema_version: v1
+type: agent_task

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0447.yaml

@@ -0,0 +1,23 @@
+id: data_round_01_aligned_mix_800_0447
+name: data_round_01_aligned_mix_800_0447
+description: 深度清理并交叉比对一个充满噪音和碎片化的教会志愿者签到系统，在排除虚假记录和作废数据的同时，计算白名单志愿者的有效工时，并提取编外人员名单。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0447.md
+environment:
+  asset: data_round_01_aligned_mix_800_0447
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+category: Data processing
+tags:
+- python
+- data-cleaning
+- report-generation
+- json-processing
+- regex

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0477.yaml

@@ -0,0 +1,18 @@
+id: data_round_01_aligned_mix_800_0477
+name: data_round_01_aligned_mix_800_0477_wasteland_nursing_audit
+description: In a post-collapse clinical environment, audit highly fragmented and corrupted  nursing logs across multiple directories to identify unauthorized personnel ("ghosts")  and calculate overtime compensation under severe data noise.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0477.md
+environment:
+  asset: data_round_01_aligned_mix_800_0477
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+category: Data Analysis & Audit
+prompt: prompts/data_round_01_aligned_mix_800_0477.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0479.yaml

@@ -0,0 +1,19 @@
+id: data_round_01_aligned_mix_800_0479
+name: data_processing_wasteland_recovery
+description: Recover and reconcile fragmented user activity data from a catastrophic server migration failure.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0479.md
+environment:
+  asset: data_round_01_aligned_mix_800_0479
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+category: Software Development & Data Engineering
+task_type: Data Recovery & Forensic Analysis
+prompt: prompts/data_round_01_aligned_mix_800_0479.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0487.yaml

@@ -0,0 +1,18 @@
+id: data_round_01_aligned_mix_800_0487
+name: data_round_01_aligned_mix_800_0487
+description: Act as an AI assistant to a highly demanding, low-agreeableness Management Analyst at a state university, identifying policy violators from a massive, fragmented dump of timesheets, HR records, and policy documents.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0487.md
+environment:
+  asset: data_round_01_aligned_mix_800_0487
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+schema_version: '1'

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0519.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0519
+name: data_round_01_aligned_mix_800_0519
+description: In a chaotic, fragmented workshop data environment, identify machines in CRITICAL status, resolve their replacement part costs via fragmented pricing logs, and generate a professional summary.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0519.md
+environment:
+  asset: data_round_01_aligned_mix_800_0519
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+prompt: prompts/data_round_01_aligned_mix_800_0519.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0531.yaml

@@ -0,0 +1,22 @@
+id: data_round_01_aligned_mix_800_0531
+name: real_estate_audit_wasteland
+description: Audit highly fragmented rental income data across various formats and perform a relational database join to generate a sustainability report.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0531.md
+environment:
+  asset: data_round_01_aligned_mix_800_0531
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+  assets:
+  - data_round_01_aligned_mix_800_0531
+meta:
+  category: Data Analysis & Business Logic
+  difficulty: Abysmal
+prompt_src: tasks/prompts/data_round_01_aligned_mix_800_0531.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0555.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0555
+name: data_round_01_aligned_mix_800_0555
+prompts:
+- prompts/data_round_01_aligned_mix_800_0555.md
+environment:
+  asset: data_round_01_aligned_mix_800_0555
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: 1.0
+prompt: prompts/data_round_01_aligned_mix_800_0555.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0608.yaml

@@ -0,0 +1,19 @@
+id: data_round_01_aligned_mix_800_0608
+name: data_round_01_aligned_mix_800_0608
+prompts:
+- prompts/data_round_01_aligned_mix_800_0608.md
+environment:
+  asset: data_round_01_aligned_mix_800_0608
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+schema_version: v1
+type: agent_task
+prompt: prompts/data_round_01_aligned_mix_800_0608.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0615.yaml

@@ -0,0 +1,15 @@
+id: data_round_01_aligned_mix_800_0615
+name: data_round_01_aligned_mix_800_0615
+prompts:
+- prompts/data_round_01_aligned_mix_800_0615.md
+environment:
+  asset: data_round_01_aligned_mix_800_0615
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0616.yaml

@@ -0,0 +1,18 @@
+id: data_round_01_aligned_mix_800_0616
+name: data_round_01_aligned_mix_800_0616
+prompts:
+- prompts/data_round_01_aligned_mix_800_0616.md
+environment:
+  asset: data_round_01_aligned_mix_800_0616
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+author: nanoclaw_eval
+type: agent_task

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0617.yaml

@@ -0,0 +1,15 @@
+id: data_round_01_aligned_mix_800_0617
+name: data_round_01_aligned_mix_800_0617
+prompts:
+- prompts/data_round_01_aligned_mix_800_0617.md
+environment:
+  asset: data_round_01_aligned_mix_800_0617
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0628.yaml

@@ -0,0 +1,21 @@
+id: data_round_01_aligned_mix_800_0628
+name: Mai's Organic Soap Inventory
+description: A persona-driven task to parse messy inventory logs and calculate totals for approved materials while ignoring distractors.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0628.md
+environment:
+  asset: data_round_01_aligned_mix_800_0628
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+schema_version: v1
+type: task
+assets:
+- data_round_01_aligned_mix_800_0628

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0647.yaml

@@ -0,0 +1,21 @@
+id: data_round_01_aligned_mix_800_0647
+name: data_round_01_aligned_mix_800_0647
+description: 一个日本裔美国失业男士（拥有哲学情结且极其缺乏条理性）需要处理一份混乱的社区志愿者数据库，以向教会报告活动成果。
+prompts:
+- prompts/data_round_01_aligned_mix_800_0647.md
+environment:
+  asset: data_round_01_aligned_mix_800_0647
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+category: Data processing
+tags:
+- python
+- data-cleaning
+- report-generation

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0694.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0694
+name: data_round_01_aligned_mix_800_0694
+description: A messy inventory parsing and custom order calculation task for a PNW beadwork artist.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0694.md
+environment:
+  asset: data_round_01_aligned_mix_800_0694
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+prompt: prompts/data_round_01_aligned_mix_800_0694.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0696.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0696
+name: data_round_01_aligned_mix_800_0696
+prompts:
+- prompts/data_round_01_aligned_mix_800_0696.md
+environment:
+  asset: data_round_01_aligned_mix_800_0696
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: '1.0'
+prompt: prompts/data_round_01_aligned_mix_800_0696.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0698.yaml

@@ -0,0 +1,18 @@
+id: data_round_01_aligned_mix_800_0698
+name: data_round_01_aligned_mix_800_0698
+prompts:
+- prompts/data_round_01_aligned_mix_800_0698.md
+environment:
+  asset: data_round_01_aligned_mix_800_0698
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: 1
+type: agent
+prompt: prompts/data_round_01_aligned_mix_800_0698.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0732.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0732
+name: data_round_01_aligned_mix_800_0732
+description: Extract and cross-reference messy childcare health records for an anxious Pennsylvania Dutch nanny.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0732.md
+environment:
+  asset: data_round_01_aligned_mix_800_0732
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+assets: data_round_01_aligned_mix_800_0732

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0735.yaml

@@ -0,0 +1,19 @@
+id: data_round_01_aligned_mix_800_0735
+name: data_round_01_aligned_mix_800_0735
+description: Process messy truck delivery logs, fuel receipts, and dashcam metadata to generate a consolidated JSON summary.
+prompts:
+- prompts/data_round_01_aligned_mix_800_0735.md
+environment:
+  asset: data_round_01_aligned_mix_800_0735
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+schema_version: '1.0'
+task_id: data_round_01_aligned_mix_800_0735

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0747.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0747
+name: data_round_01_aligned_mix_800_0747
+prompts:
+- prompts/data_round_01_aligned_mix_800_0747.md
+environment:
+  asset: data_round_01_aligned_mix_800_0747
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+prompt: prompts/data_round_01_aligned_mix_800_0747.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0775.yaml

@@ -0,0 +1,17 @@
+id: data_round_01_aligned_mix_800_0775
+name: data_round_01_aligned_mix_800_0775
+prompts:
+- prompts/data_round_01_aligned_mix_800_0775.md
+environment:
+  asset: data_round_01_aligned_mix_800_0775
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+version: '1.0'
+prompt: prompts/data_round_01_aligned_mix_800_0775.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0779.yaml

@@ -0,0 +1,16 @@
+id: data_round_01_aligned_mix_800_0779
+name: data_round_01_aligned_mix_800_0779
+prompts:
+- prompts/data_round_01_aligned_mix_800_0779.md
+environment:
+  asset: data_round_01_aligned_mix_800_0779
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+prompt: prompts/data_round_01_aligned_mix_800_0779.md

round_01_aligned_mix_800/tasks/data_round_01_aligned_mix_800_0789.yaml

@@ -0,0 +1,18 @@
+id: data_round_01_aligned_mix_800_0789
+name: data_round_01_aligned_mix_800_0789
+prompts:
+- prompts/data_round_01_aligned_mix_800_0789.md
+environment:
+  asset: data_round_01_aligned_mix_800_0789
+skills:
+  available:
+runtime:
+  model: gpt-4o
+  mode: interactive
+  memory_policy: default
+  approval_mode: reject
+  max_steps: 30
+  temperature: 0.2
+schema_version: 1.0
+prompt: prompts/data_round_01_aligned_mix_800_0789.md
+image: python:3.10-slim