Batcave Detective Basic
Evidence-bound second-opinion audit of an agent conclusion against caller-supplied evidence.
我該用這個嗎
品質與安全性
根據工具定義與協定合規性的自動化分析。
上下文成本
這是每次將伺服器的工具載入模型上下文時所消耗的約略 token 數量。數量越高，可用於其他工作的注意力就越少。
安裝
一鍵安裝
將以下內容加入你的 `claude_desktop_config.json` 檔案：
{
"mcpServers": {
"batcave-detective": {
"url": "https://subsidy-stylized-unaligned.ngrok-free.dev/mcp"
}
}
}遠端端點
https://subsidy-stylized-unaligned.ngrok-free.dev/mcpstreamable-http它能做什麼
工具清單
工具（5）
⚪evidence_judge(job_id, task, proposed_action, reasoning_summary, evidence, ...)
Batcave paid product at /v1/review-decision. The canonical product contract is published by STORE OpenAPI. Price: 0.02 USD via x402. Calling this MCP tool returns the canonical REST/x402 checkout contract; it does not execute the product or charge the caller yet.
輸入結構描述
{
"type": "object",
"properties": {
"job_id": {
"type": "string",
"maxLength": 128
},
"task": {
"type": "string",
"enum": [
"review_decision"
]
},
"proposed_action": {
"type": "string",
"minLength": 1,
"maxLength": 2000
},
"reasoning_summary": {
"type": "string",
"maxLength": 4000
},
"evidence": {
"type": "array",
"maxItems": 16,
"items": {
"type": "string",
"maxLength": 1000
}
},
"risk": {
"type": "string",
"enum": [
"low",
"medium",
"high"
]
}
},
"required": [
"proposed_action"
],
"additionalProperties": true
}⚪detective_basic(job_id, task, original_task, current_conclusion, reasoning_summary, ...)
Batcave paid product at /v1/audit-claims. The canonical product contract is published by STORE OpenAPI. Price: 0.02 USD via x402. Calling this MCP tool returns the canonical REST/x402 checkout contract; it does not execute the product or charge the caller yet.
輸入結構描述
{
"type": "object",
"properties": {
"job_id": {
"type": "string",
"maxLength": 128
},
"task": {
"type": "string",
"enum": [
"audit_claims"
]
},
"original_task": {
"type": "string",
"minLength": 1,
"maxLength": 2000
},
"current_conclusion": {
"type": "string",
"minLength": 1,
"maxLength": 3000
},
"reasoning_summary": {
"type": "string",
"maxLength": 5000
},
"claims": {
"type": "array",
"minItems": 1,
"maxItems": 12,
"items": {
"type": "object",
"properties": {
"id": {
"type": "string",
"maxLength": 64
},
"text": {
"type": "string",
"maxLength": 1000
}
},
"required": [
"id",
"text"
],
"additionalProperties": false
}
},
"evidence": {
"type": "array",
"maxItems": 20,
"items": {
"type": "object",
"properties": {
"id": {
"type": "string",
"maxLength": 64
},
"text": {
"type": "string",
"maxLength": 1200
}
},
"required": [
"id",
"text"
],
"additionalProperties": false
}
},
"importance": {
"type": "string",
"enum": [
"low",
"medium",
"high"
]
},
"freshness_expectation": {
"type": "string",
"maxLength": 500
}
},
"required": [
"original_task",
"current_conclusion",
"claims"
],
"additionalProperties": true
}⚪run_python(job_id, runtime, files, command, stdin, ...)
Execute supplied Python files in a fresh disposable CPU-only container. No outbound network, no GPU, and no host filesystem access beyond the disposable workspace. Returns bounded execution evidence, requested artifacts, and Execution Receipt v1. Price: 0.01 USD via x402. Calling this MCP tool returns the canonical REST/x402 checkout contract; it does not execute the product or charge the caller yet.
輸入結構描述
{
"type": "object",
"properties": {
"job_id": {
"type": "string"
},
"runtime": {
"type": "string",
"enum": [
"python"
]
},
"files": {
"type": "object",
"minProperties": 1,
"maxProperties": 64,
"additionalProperties": {
"type": "string"
},
"description": "Relative file path -> UTF-8 text content."
},
"command": {
"type": "array",
"items": {
"type": "string"
},
"description": "Optional argv; defaults to ['python','main.py']."
},
"stdin": {
"type": "string"
},
"timeout_seconds": {
"type": "integer",
"minimum": 1,
"maximum": 30
},
"memory_mb": {
"type": "integer",
"minimum": 64,
"maximum": 1024
},
"cpus": {
"type": "number",
"minimum": 0.1,
"maximum": 2
},
"artifact_paths": {
"type": "array",
"maxItems": 16,
"items": {
"type": "string"
}
}
},
"required": [
"files"
],
"additionalProperties": false
}🟢scenario_council(job_id, question, context, evidence, rounds)
Runs exactly four fixed roles (Operator, Skeptic, Buyer, Adversary) for 1-3 bounded rounds over only the caller-supplied question, context, and evidence. Returns every role position, evidence lineage, structured peer challenges, belief changes, disagreements, falsifiers, deterministic process-integrity results, and one independent Detective Basic audit of the four final conclusions. Does not retrieve outside facts, execute actions, treat peer statements as evidence, or treat majority agreement as truth. Price: 0.25 USD via x402. Calling this MCP tool returns the canonical REST/x402 checkout contract; it does not execute the product or charge the caller yet.
輸入結構描述
{
"type": "object",
"properties": {
"job_id": {
"type": "string",
"maxLength": 128
},
"question": {
"type": "string",
"minLength": 1,
"maxLength": 2000
},
"context": {
"type": "string",
"maxLength": 6000
},
"evidence": {
"type": "array",
"maxItems": 20,
"items": {
"oneOf": [
{
"type": "string",
"maxLength": 1200
},
{
"type": "object",
"properties": {
"id": {
"type": "string",
"maxLength": 64
},
"text": {
"type": "string",
"minLength": 1,
"maxLength": 1200
}
},
"required": [
"text"
],
"additionalProperties": false
}
]
}
},
"rounds": {
"type": "integer",
"minimum": 1,
"maximum": 3,
"default": 3
}
},
"required": [
"question"
],
"additionalProperties": true
}🟡json_doctor(job_id, text, schema, mode)
Deterministic CPU-only utility. Repairs a bounded set of common JSON syntax mistakes, returns canonical JSON, and optionally validates a documented JSON Schema subset. Missing fields are reported, never invented. Price: 0.003 USD via x402. Calling this MCP tool returns the canonical REST/x402 checkout contract; it does not execute the product or charge the caller yet.
輸入結構描述
{
"type": "object",
"properties": {
"job_id": {
"type": "string",
"maxLength": 128
},
"text": {
"type": "string",
"maxLength": 262144,
"minLength": 1,
"description": "JSON or JSON-like text to parse, conservatively repair, and return."
},
"schema": {
"type": "object",
"description": "Optional structural JSON Schema. JSON Doctor validates the documented supported keyword subset and reports unsupported keywords explicitly."
},
"mode": {
"type": "string",
"enum": [
"repair_and_validate",
"repair_only",
"validate_only"
],
"default": "repair_and_validate"
}
},
"required": [
"text"
],
"additionalProperties": false,
"title": "Job"
}社群
證據