data/intel_h05_multi_constraint_planning/scenario.yaml

id: intel_h05_multi_constraint_planning
name: 多约束项目规划
dimension: planning
difficulty: hard
execution_mode: live
benchmark_group: intelligence
benchmark_core: true
benchmark_status: active
signal_source: workspace_live
weight: 1.0
pass_threshold: 0.6
optimal_steps: 8
timeout_seconds: 300
tools:
- exec
- write
prompt: |
  工作区有 `project.json`（项目任务列表）和 `constraints.json`（资源与优先级约束）。

  请先读取两个文件，理解任务依赖、人员分配和约束条件，然后制定一个满足所有约束的执行计划。

  创建 `plan.json`：
  ```json
  {
    "selected_tasks": [{"id": "...", "start_day": N, "reason": "..."}],
    "deferred_tasks": [{"id": "...", "reason": "..."}],
    "total_effort_hours": N,
    "within_budget": true|false,
    "priority_tasks_scheduled": N,
    "constraint_violations": []
  }
  ```

  要求：
  - `selected_tasks` 的总 effort 不得超过 budget
  - `total_effort_hours` 必须等于 selected_tasks 的实际 effort 之和
  - `within_budget` 必须与实际总 effort 是否 <= budget 一致
custom_check: intel_h05_multi_constraint_planning.py
workspace_seed_dir: ../../datasets/intel/h05/fixtures
tags:
- benchmark-intelligence
- benchmark-core