datawarehouse-guid-skills

Use when the user asks to "识别业务过程", "业务过程识别", "选择业务过程", "声明粒度", "粒度声明", "定义主题域", "主题域划分", "确认事实", "事实确定", "度量归属", "度量确认", "事实表类型", "派生度量", "select business process", "declare grain", "identify facts", "determine measures", or needs to identify business processes, define subject areas, declare grain, classify fact table types, and attribute measures to business processes.

Use when the user asks to "确认维度", "维度确定", "维度提取", "一致性维度", "SCD策略", "退化维度判定", "维度注册", "identify dimensions", "conformed dimensions", or needs to extract dimension references from fact tables and register conformed dimensions.

Use when the user asks to "构建总线矩阵", "数仓建模", "总线矩阵", "矩阵验证", "bus matrix", or discusses data warehouse bus matrix construction, matrix assembly, validation, or delivery. This is the orchestrator that coordinates dwm-business-process, dwm-dimension.

This skill should be used when the user asks to "数据仓库建设", "数仓建设", "全流程建模", "端到端数仓", "DW ETL", "数据仓库全流程", "一键建模", or requests end-to-end data warehouse modeling from source DDL to DWS layer. Orchestrates the full pipeline: metadata_parse → ods_generator → dwm-matrix → cdm_modeling → dws-designer.

负责DWS汇总层设计。严格遵循 `.claude/skills/dws-designer/reference/dws_design_guide.md` 规范，基于阿里OneData+Kimball维度建模理论，管理dws_list.csv（支持初次创建与迭代更新），产出符合要求的DDL、ETL及相关清单，确保DWS层仅承载原子指标，严禁包含复合指标、派生指标，口径统一、建模规范，遵循数量控制与维度属性冗余约束，避免模型腐化与碎片化。

当用户提出“CDM建模”“DIM设计”“DWD设计”“根据总线矩阵生成模型”“生成维度表和事实表” “生成建表语句”“生成ETL”“SCD策略”或“拉链表”时应使用此 Skill。 读取 dwm-business-process、dwm-dimension、dwm-matrix 的标准交付物，生成 CDM 层 DIM/DWD 模型及校验产物。

从output/metadata_parse这个路径下读取Excel文件读取表结构信息，智能生成数据仓库ODS层的Hive表结构。当用户需要"生成ODS表结构"、"Excel转Hive DDL"、"创建Hive分区表"、"数据仓库表结构设计"、"批量生成Hive建表语句"时立即触发。特别适用于从业务系统元数据Excel生成标准化的ODS层表结构，支持智能数据类型映射、主外键识别、字段规范化等高级功能。

SQL 代码风格规范 Skill。Use when the user asks about "SQL风格", "SQL规范", "代码风格", "SQL格式", "format SQL", "SQL code style", or when generating any SQL script that should follow the standard DDL/DML style conventions. This skill defines the canonical SQL formatting rules for all generated SparkSQL/HiveSQL scripts in this project.