/alibabacloud-lakehouse-migration

name: alibabacloud-lakehouse-migration
version: 1.0.0
description: Unified dispatch entry point for the LHM (Lakehouse Migration Center) skill, coordinating the three core capabilities of SQL conversion, data validation, and big-data workflow migration. Use when the user raises a lakehouse migration requirement, such as SQL dialect conversion (e.g., "presto to spark", "hive to maxcompute", "转换 SQL", "SQL 转换", "方言转换"), pre/post-migration data validation / data consistency comparison ("数据校验", "数据比对", "一致性检查", "跑校验", "数据量校验", "指标校验"), migration of scheduled workflows such as DataWorks/DolphinScheduler/Airflow ("工作流迁移", "调度任务迁移", "建数据源", "创建数据源", "把 X 迁移到 Y"), or when the user is unsure which sub-skill to use or needs cross-skill collaboration.

LHM Lakehouse Migration Unified Dispatcher

This skill is the unified entry point for the three core skills of Alibaba Cloud Lakehouse Migration Center (LHM). It analyzes user intent and intelligently routes requests to the corresponding sub-skill for execution.

Supported Sub-Skills

| Skill | Description | Trigger Scenario | |-------|-------------|------------------| | **SQL Conversion** | Converts SQL from a source dialect to a target analytics-engine dialect | User provides SQL text or a directory and specifies source/target dialects | | **Data Validation** | Validates data consistency across heterogeneous data sources | User needs to compare data before/after migration or verify data integrity | | **Big-Data Workflow Migration** | Migrates scheduled jobs, environment configurations, and deployment execution | User needs to migrate DataWorks/MaxCompute or similar workflows |

> All three sub-skills are peer-level packages that live alongside this dispatcher in the same monorepo. In a monorepo deployment they are accessed by **absolute file path** under `${SKILL_HOME}`; when installed as independent Skill packages they can also be invoked by **skill name**. The Agent must try both approaches (path first, then name) before concluding a sub-skill is unavailable. > > ⚠️ **Discovery boundary**: skill discovery registers only the top-level SKILL.md of each installed package. Files nested inside a package (e.g., `skill/lhm-sch-*/SKILL.md`, `skills/sql-atomic-conversion/SKILL.md`, `atomic-skills/lhm-*/SKILL.md`) are **NOT registered as invokable skills** — they can only be loaded by file path from within their parent package. Never route to them with "Invoke skill: ..."; route to them by absolute file path resolved from the parent package's `SKILL_HOME`.

Path Resolution (SKILL_HOME)

`${SKILL_HOME}` appears throughout this file. It is **not** a pre-set environment variable — the Agent must resolve it before running any command. `SKILL_HOME` is the absolute path of the directory containing this SKILL.md:

# If the load path of this SKILL.md is known, take its containing directory:
# SKILL_HOME = <absolute path of the directory containing this SKILL.md>
echo $SKILL_HOME  # confirm the path is correct before any command

All subsequent `${SKILL_HOME}/...` paths must be expanded to this absolute path before execution; never leave the literal `${SKILL_HOME}` token in a shell command.

**Sub-skill location**: all three sub-skill packages reside under the `references/` subdirectory within `SKILL_HOME`. The canonical path for a sub-skill's SKILL.md is:

${SKILL_HOME}/references/<sub-skill-dir>/SKILL.md

For example:

• SQL Conversion: `${SKILL_HOME}/references/lhm-sql-conversion/SKILL.md`
• Data Validation: `${SKILL_HOME}/references/lhm-data-validation-skill/SKILL.md`
• Workflow Migration: `${SKILL_HOME}/references/lhm-bigdata-workflow-migration-skill/SKILL.md`

**Sub-agent dispatch rule**: when delegating a task to a sub-agent (Task/subagent mechanism of the current platform), the task description must contain **fully resolved absolute paths** — expand every `${SKILL_HOME}` before dispatching. A sub-agent runs in an independent context and cannot resolve this variable on its own.

Intent Recognition and Routing Rules

1. SQL Conversion Skill (`lhm-sql-conversion`)

**Trigger keywords**:

• "convert SQL", "SQL conversion", "dialect conversion" (the Chinese equivalents are declared in the frontmatter description)
• "presto to spark", "hive to maxcompute", "clickhouse to hologres"
• SQL text provided together with source/target parameters
• A directory path containing .sql files

**Routing action**:

Invoke skill: lhm-sql-conversion

**Typical user inputs**:

• "Convert this presto SQL to spark format"
• "Batch-convert all hive SQL under /path/to/sqls to maxcompute"
• "Convert SELECT format_datetime(now(), 'yyyy-MM-dd') to spark"

---

2. Data Validation Skill (`lhm-data-validation-skill`)

**Trigger keywords**:

• "data validation", "data comparison", "consistency check"
• "run validation", "verify data", "check data"
• "count check", "metric check", "content check"
• Mentions of the `alibabacloud_lhm` SDK, ds_id, or batch_id

**Routing action**:

Invoke skill: lhm-data-validation-skill

**Typical user inputs**:

• "Run the count check between source MySQL and target MaxCompute"
• "View the result report of the last validation"
• "Rerun the failed validation tasks"
• "Diagnose why these two tables have inconsistent data"

---

3. Big-Data Workflow Migration Skill (`lhm-bigdata-workflow-migration-skill`)

This skill provides three core capabilities:

| Capability | Trigger | Action | |------------|---------|--------| | **Query skill capabilities** | User asks "what can this skill do / introduce its capabilities" | Output the content of `docs/introduction.md` | | **Create data source** | Instructions such as "create a data source of type XX ..." | Route to `skill/lhm-sch-ds/SKILL.md` to complete data source creation | | **Schedule migration** | A migration instruction (e.g., "migrate X to Y") | Config validation → environment initialization → data source validation → exploration & conversion → deployment &