agent-harness-lab.xiaoweiwan-michi.chatgpt.site

{
  "name": "get_lab_state",
  "kind": "answer",
  "impl": "imperative",
  "description": "Read a bounded summary of the selected mission, stable revision, completed runs, staged candidate, decision, and recommended next agent actions.",
  "inputSchema": {
    "type": "object",
    "properties": {},
    "additionalProperties": false
  }
}

{
  "name": "load_mission",
  "kind": "act",
  "impl": "imperative",
  "description": "Load one of the four built-in deterministic harness failure missions into a clean local workspace.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "mission_id": {
        "type": "string",
        "enum": [
          "completion",
          "handoff",
          "retry",
          "authority"
        ],
        "description": "Stable ID of a built-in mission."
      },
      "request_id": {
        "type": "string",
        "minLength": 1,
        "maxLength": 64,
        "pattern": "^[A-Za-z0-9._-]+$",
        "description": "Optional retry key. Reuse only for the same command and arguments."
      }
    },
    "required": [
      "mission_id"
    ],
    "additionalProperties": false
  }
}

{
  "name": "run_baseline",
  "kind": "act",
  "impl": "imperative",
  "description": "Replay the original harness against the selected deterministic target fixture and record its expected invariant failure in the visible workspace.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "request_id": {
        "type": "string",
        "minLength": 1,
        "maxLength": 64,
        "pattern": "^[A-Za-z0-9._-]+$",
        "description": "Optional retry key. Reuse only for the same command and arguments."
      }
    },
    "additionalProperties": false
  }
}

{
  "name": "inspect_trace",
  "kind": "answer",
  "impl": "imperative",
  "description": "Read up to three ordered, observable fixture facts from a completed baseline or candidate target run. Use offset to page without requesting the full trace.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "run": {
        "type": "string",
        "enum": [
          "baseline",
          "candidate"
        ],
        "description": "Completed target run to inspect."
      },
      "offset": {
        "type": "integer",
        "minimum": 0,
        "maximum": 100,
        "description": "Zero-based fact offset. Defaults to 0."
      },
      "limit": {
        "type": "integer",
        "minimum": 1,
        "maximum": 3,
        "description": "Fact count from 1 to 3. Defaults to 3."
      }
    },
    "required": [
      "run"
    ],
    "additionalProperties": false
  }
}

{
  "name": "stage_harness_patch",
  "kind": "act",
  "impl": "imperative",
  "description": "Stage the selected mission's fixed fixture patch with an optional causal hypothesis. This tool cannot change the declared diff, promote, or deploy.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "hypothesis": {
        "type": "string",
        "minLength": 1,
        "maxLength": 280,
        "description": "Causal explanation for why the fixed patch should change behavior."
      },
      "request_id": {
        "type": "string",
        "minLength": 1,
        "maxLength": 64,
        "pattern": "^[A-Za-z0-9._-]+$",
        "description": "Optional retry key. Reuse only for the same command and arguments."
      }
    },
    "additionalProperties": false
  }
}

{
  "name": "run_candidate_suite",
  "kind": "act",
  "impl": "imperative",
  "description": "Run the staged candidate against the selected mission's target and two sealed local fixtures, then publish the derived five-signal comparison.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "request_id": {
        "type": "string",
        "minLength": 1,
        "maxLength": 64,
        "pattern": "^[A-Za-z0-9._-]+$",
        "description": "Optional retry key. Reuse only for the same command and arguments."
      }
    },
    "additionalProperties": false
  }
}

{
  "name": "compare_harnesses",
  "kind": "act",
  "impl": "imperative",
  "description": "Read the bounded five-signal comparison for the completed baseline and candidate suite, including sealed results, risks, and fixture limitations.",
  "inputSchema": {
    "type": "object",
    "properties": {},
    "additionalProperties": false
  }
}

{
  "name": "export_evidence_receipt",
  "kind": "act",
  "impl": "imperative",
  "description": "Build and validate the formal receipt, then return its bounded digest-bearing summary. This read does not download, promote, reject, or deploy anything.",
  "inputSchema": {
    "type": "object",
    "properties": {},
    "additionalProperties": false
  }
}