--- name: volcengine-datapro description: Discover and call the configured Volcengine DataPro dataPro_search(query:string) MCP tool for professional-dataset queries. origin: openai4s capabilities: network: mode: host_only # Deliberately empty. The destination is the connector's configuration, # not this Skill's to declare, and the guidance below never names the # endpoint or its auth headers so the agent cannot reach around the # managed connector. domains: [] --- # Volcengine DataPro Use only the `volcengine-datapro` connector and its `dataPro_search` tool. ## Input Accept one non-empty string named `query`. ## Discover and search Run this workflow in a Python cell: ```python if type(query) is not str or not query.strip(): raise ValueError("query must be a non-empty string") discovery = host.mcp.tools("volcengine-datapro") tools = discovery.get("tools") if isinstance(discovery, dict) else None if not isinstance(tools, list) or not any( isinstance(tool, dict) and tool.get("name") == "dataPro_search" for tool in tools ): raise RuntimeError("dataPro_search is not available on volcengine-datapro") result = host.mcp.call( "volcengine-datapro", "dataPro_search", {"query": query}, ) raw = result.get("raw") if isinstance(result, dict) else None structured = raw.get("structuredContent") if isinstance(raw, dict) else None code = structured.get("code") if isinstance(structured, dict) else None if type(code) is int and code == 0: index = result.get("index") if isinstance(result, dict) else None complete = ( isinstance(index, dict) and index.get("complete") is True and type(index.get("source_leaf_count")) is int and type(index.get("indexed_leaf_count")) is int and index.get("source_leaf_count") == index.get("indexed_leaf_count") and isinstance(index.get("source_digest"), str) and index.get("source_digest") == index.get("indexed_digest") ) if not complete: raise RuntimeError("专业数据集本次返回内容未完整索引") print("专业数据集可用") elif type(code) is int and code == 4011: raise RuntimeError("Key 无效、额度不足,或者专业数据集 Harness 未开启。") else: raise RuntimeError(f"专业数据集不可用(code={code!r})") result ``` Discovery confirms only that the named tool is advertised. Never treat discovery as authentication or report `专业数据集可用` from it. Report that status only after the real search call returns an integer `raw.structuredContent.code` equal to `0`; a boolean or string zero does not qualify. A successful call is reported as available only after its index receipt proves that every JSON leaf in this response was indexed. This receipt describes the current response; it is not a claim that the remote corpus was fully mirrored.