📖 技能介绍

Data Migration Planner

Plan, execute, and validate data migrations between systems. Covers schema mapping, ETL pipeline design, rollback strategies, and post-migration validation.

What It Does

Given source and target system details, this skill: 1. Maps source → target schemas with field-level transformation rules 2. Generates an ETL pipeline plan with staging, transform, and load phases 3. Creates validation queries (row counts, checksum, referential integrity) 4. Builds a rollback plan with point-of-no-return criteria 5. Produces a migration runbook with go/no-go gates

Usage

Tell your agent: - "Plan a migration from Salesforce to HubSpot CRM" - "Create a data migration runbook for moving from MySQL to PostgreSQL" - "Map our legacy ERP data to the new system schema"

Migration Framework

Phase 1: Discovery

Inventory all source tables/objects and record counts
Document data types, constraints, and relationships
Identify data quality issues (nulls, duplicates, orphans)
Map business rules that affect data interpretation

Phase 2: Schema Mapping

For each source entity, document: | Source Field | Type | Target Field | Type | Transform | Notes | |---|---|---|---|---|---| | (field) | (type) | (field) | (type) | (rule) | (edge cases) |

Phase 3: ETL Pipeline

Extract → Stage (raw) → Clean → Transform → Validate → Load → Verify

Extract: Full vs incremental, API vs direct DB, rate limits
Stage: Raw landing zone, no transforms, audit trail
Clean: Dedup, null handling, encoding fixes
Transform: Type conversions, lookups, calculated fields
Validate: Pre-load checks (counts, checksums, business rules)
Load: Batch size, parallelism, error handling
Verify: Post-load reconciliation

Phase 4: Validation

Row count match (source vs target, per table)
Checksum validation on key columns
Referential integrity checks
Business rule validation (e.g., all active accounts migrated)
User acceptance sampling (random 5% manual review)

Phase 5: Cutover

Go/no-go criteria checklist
Point-of-no-return definition
Rollback procedure and time estimate
Communication plan (users, stakeholders)
Parallel run period (if applicable)

Risk Factors

Data volume: >10M rows = batch strategy required
Downtime window: Zero-downtime needs CDC/dual-write
Data quality: Garbage in = garbage out. Clean BEFORE migrating
Dependencies: Other systems reading from source during migration
Compliance: GDPR/HIPAA data handling during transit

Output Format

Deliver a migration runbook as structured markdown with: 1. Executive summary (what, why, when, risk level) 2. Schema mapping tables 3. ETL pipeline specification 4. Validation test suite 5. Cutover runbook with rollback 6. Timeline with milestones

Cost Estimation

Typical migration costs by complexity: - Simple (1-5 tables, <1M rows): $5K-$15K or 1-2 weeks internal - Medium (10-50 tables, 1-10M rows): $25K-$75K or 1-2 months - Complex (50+ tables, 10M+ rows, multiple systems): $100K-$500K or 3-6 months

Built by AfrexAI — AI Context Packs for business automation.

Calculate your AI automation ROI: Revenue Calculator

🤖 AI 评测

这个 Skill 质量不错，框架设计清晰，能帮助用户从整体上理解和规划数据迁移的各个阶段。优点是结构完整、风险考虑全面、使用方式简单明了，用户通过一句话就能启动迁移规划。但不足也很明显：缺少具体的操作模板和工具，用户拿到规划后可能仍不知道如何下手；也没有针对不同难度项目的具体指导；另外成本估算信息比较简略。总体来说，这是一个好的规划框架，但在落地执行层面还有提升空间。