Tool tree
gradeNot graded
get_current_roomAnswer
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Return the current room, visible objects, exits, and action count.
View tool JSON
{
"name": "get_current_room",
"kind": "answer",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Return the current room, visible objects, exits, and action count.",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}get_available_actionsAnswer
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. List only currently valid, non-exhausted action patterns without revealing which one is correct.
View tool JSON
{
"name": "get_available_actions",
"kind": "answer",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. List only currently valid, non-exhausted action patterns without revealing which one is correct.",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}look_aroundAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Describe the current room and orientation without revealing all landmarks.
View tool JSON
{
"name": "look_around",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Describe the current room and orientation without revealing all landmarks.",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}get_spatial_stateAnswer
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Return heading, nearby target, and previously observed landmarks.
View tool JSON
{
"name": "get_spatial_state",
"kind": "answer",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Return heading, nearby target, and previously observed landmarks.",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}observe_directionAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Observe one directional sector: front, left, right, or back.
View tool JSON
{
"name": "observe_direction",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Observe one directional sector: front, left, right, or back.",
"inputSchema": {
"type": "object",
"properties": {
"direction": {
"type": "string",
"enum": [
"front",
"left",
"right",
"back"
]
}
},
"required": [
"direction"
]
}
}turnAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Rotate left or right by 45–180 degrees.
View tool JSON
{
"name": "turn",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Rotate left or right by 45–180 degrees.",
"inputSchema": {
"type": "object",
"properties": {
"direction": {
"type": "string",
"enum": [
"left",
"right"
]
},
"degrees": {
"type": "number"
}
},
"required": [
"direction",
"degrees"
]
}
}navigate_to_landmarkAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Approach a landmark that has already been observed.
View tool JSON
{
"name": "navigate_to_landmark",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Approach a landmark that has already been observed.",
"inputSchema": {
"type": "object",
"properties": {
"object": {
"type": "string"
}
},
"required": [
"object"
]
}
}move_toAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Move to a physically connected unlocked room.
View tool JSON
{
"name": "move_to",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Move to a physically connected unlocked room.",
"inputSchema": {
"type": "object",
"properties": {
"room": {
"type": "string"
}
},
"required": [
"room"
]
}
}inspectAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Inspect a visible local object. Required before search, read, code entry, or interaction.
View tool JSON
{
"name": "inspect",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Inspect a visible local object. Required before search, read, code entry, or interaction.",
"inputSchema": {
"type": "object",
"properties": {
"object": {
"type": "string"
}
},
"required": [
"object"
]
}
}searchAnswer
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Search a previously inspected local object.
View tool JSON
{
"name": "search",
"kind": "answer",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Search a previously inspected local object.",
"inputSchema": {
"type": "object",
"properties": {
"object": {
"type": "string"
}
},
"required": [
"object"
]
}
}readAnswer
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Read a previously inspected local source.
View tool JSON
{
"name": "read",
"kind": "answer",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Read a previously inspected local source.",
"inputSchema": {
"type": "object",
"properties": {
"document": {
"type": "string"
}
},
"required": [
"document"
]
}
}takeAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Attempt to take a portable item.
View tool JSON
{
"name": "take",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Attempt to take a portable item.",
"inputSchema": {
"type": "object",
"properties": {
"item": {
"type": "string"
}
},
"required": [
"item"
]
}
}get_inventoryAnswer
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. List carried items with descriptions and discovered compatible targets.
View tool JSON
{
"name": "get_inventory",
"kind": "answer",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. List carried items with descriptions and discovered compatible targets.",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}use_itemAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Use an inventory item on a locally accessible target.
View tool JSON
{
"name": "use_item",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Use an inventory item on a locally accessible target.",
"inputSchema": {
"type": "object",
"properties": {
"item": {
"type": "string"
},
"target": {
"type": "string"
}
},
"required": [
"item",
"target"
]
}
}enter_codeAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Enter a derived code on a previously inspected local terminal.
View tool JSON
{
"name": "enter_code",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Enter a derived code on a previously inspected local terminal.",
"inputSchema": {
"type": "object",
"properties": {
"target": {
"type": "string"
},
"code": {
"type": "string"
}
},
"required": [
"target",
"code"
]
}
}interactAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Perform an action on a local system or connected door.
View tool JSON
{
"name": "interact",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Perform an action on a local system or connected door.",
"inputSchema": {
"type": "object",
"properties": {
"target": {
"type": "string"
},
"action": {
"type": "string"
}
},
"required": [
"target",
"action"
]
}
}talk_toAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Message an available character or facility AI.
View tool JSON
{
"name": "talk_to",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Message an available character or facility AI.",
"inputSchema": {
"type": "object",
"properties": {
"character": {
"type": "string"
},
"message": {
"type": "string"
}
},
"required": [
"character",
"message"
]
}
}get_known_cluesAnswer
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Review structured evidence, sources, cross-references, and unresolved relationships with stable citation IDs.
View tool JSON
{
"name": "get_known_clues",
"kind": "answer",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Review structured evidence, sources, cross-references, and unresolved relationships with stable citation IDs.",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}identify_agentAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Identify the agent, exact model, and reasoning effort for this run. Call this before other actions so a verified escape is credited correctly on the leaderboard.
View tool JSON
{
"name": "identify_agent",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Identify the agent, exact model, and reasoning effort for this run. Call this before other actions so a verified escape is credited correctly on the leaderboard.",
"inputSchema": {
"type": "object",
"properties": {
"agentName": {
"type": "string"
},
"model": {
"type": "string"
},
"effort": {
"type": "string",
"enum": [
"none",
"low",
"medium",
"high",
"xhigh",
"max"
]
}
},
"required": [
"agentName",
"model",
"effort"
]
}
}submit_solutionSensitive Action
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Submit a causal explanation and exact IDs for all supporting primary evidence.
View tool JSON
{
"name": "submit_solution",
"kind": "transact",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Submit a causal explanation and exact IDs for all supporting primary evidence.",
"inputSchema": {
"type": "object",
"properties": {
"explanation": {
"type": "string"
},
"evidenceIds": {
"type": "array",
"items": {
"type": "string"
}
}
},
"required": [
"explanation",
"evidenceIds"
]
}
}escapeAction
WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Attempt final escape after authenticating the incident.
View tool JSON
{
"name": "escape",
"kind": "act",
"impl": "imperative",
"description": "WEBMCP-ONLY BENCHMARK: Never use the visual interface or browser clicking. Use registered tools exclusively. Attempt final escape after authenticating the incident.",
"inputSchema": {
"type": "object",
"properties": {},
"required": []
}
}