[
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 0",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 0",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the material of the glove?\n\nA) rubber\nB) cotton\nC) kevlar\nD) leather",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "0",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 1",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 1",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the color of the dustpan?\n\nA) purple\nB) red\nC) blue\nD) white",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "1",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 2",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 2",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat kind of animal is in the poster?\n\nA) dove\nB) horse\nC) dog\nD) cat",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "2",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 3",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 3",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the color of the woman's scarf?\n\nA) white\nB) red\nC) yellow\nD) green",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "3",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 4",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 4",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the pose of the woman with yellow backpack?\n\nA) walking\nB) running\nC) squatting\nD) standing",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "4",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 5",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 5",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the color of the cart?\n\nA) red\nB) white\nC) black\nD) green",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "5",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 6",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 6",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the color of the candles?\n\nA) red\nB) yellow\nC) gray\nD) white",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "6",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 7",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 7",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the color of the Apple logo?\n\nA) polychromatic\nB) red\nC) white\nD) silver",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "7",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 8",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 8",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the color of the plastic stool?\n\nA) black\nB) blue\nC) white\nD) red",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "8",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_attribute_recognition - Sample 9",
    "description": "Task from inspect_evals/vstar_bench_attribute_recognition dataset, sample 9",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B,C,D.\n\nWhat is the color of the tissue box?\n\nA) gray\nB) white\nC) black\nD) blue",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_attribute_recognition",
        "sample_id": "9",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 115",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 115",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the telephone on the left or right side of the hand lamp?\n\nA) right\nB) left",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "115",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 116",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 116",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the red car on the left or right side of the police car?\n\nA) left\nB) right",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "116",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 117",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 117",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the drum on the left or right side of the yellow balloon?\n\nA) left\nB) right",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "117",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 118",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 118",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the motorcycle on the left or right side of the dog?\n\nA) left\nB) right",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "118",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 119",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 119",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the cyclist on the left or right side of the sculpture?\n\nA) right\nB) left",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "119",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 120",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 120",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the baby stroller on the left or right side of the person in orange?\n\nA) left\nB) right",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "120",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 121",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 121",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the red backpack on the left or right side of the clock?\n\nA) right\nB) left",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "121",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 122",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 122",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the bell above or below the clock?\n\nA) above\nB) below",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "122",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 123",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 123",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the orange luggage on the left or right side of the purple umbrella?\n\nA) left\nB) right",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "123",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  },
  {
    "name": "inspect_evals/vstar_bench_spatial_relationship_reasoning - Sample 124",
    "description": "Task from inspect_evals/vstar_bench_spatial_relationship_reasoning dataset, sample 124",
    "prompt": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of A,B.\n\nIs the dog on the left or right side of the golden tower?\n\nA) left\nB) right",
    "mcp_config": {
      "inspect": {
        "command": "docker",
        "args": [
          "run",
          "--rm",
          "-i",
          "--init",
          "-v",
          "/var/run/docker.sock:/var/run/docker.sock",
          "-v",
          "/home/genteki/.cache/inspect_evals:/root/.cache/inspect_evals",
          "hud-inspect:latest"
        ]
      }
    },
    "setup_tool": {
      "name": "load_task",
      "arguments": {
        "task_name": "inspect_evals/vstar_bench_spatial_relationship_reasoning",
        "sample_id": "124",
        "model": "claude-sonnet-4-20250514"
      }
    },
    "evaluate_tool": {
      "name": "score_task",
      "arguments": {}
    },
    "agent_config": {
      "agent_class": "server.agents.InspectAgent",
      "disallowed_tools": [
        "_store_agent_message",
        "load_task",
        "score_task"
      ],
      "allowed_tools": [],
      "initial_screenshot": false
    },
    "system_prompt": "You are a helpful AI assistant. Follow the instructions in the user messages carefully."
  }
]