Data Architecture Information Generator

Problem to Solve

Data-logic modeling from MBU table-document sets often involves manually locating inputs and outputs, extracting fields, identifying business objects, designing logical models, and validating normal forms and coverage. This can miss fields, duplicate definitions, or let Markdown reports drift from source data. The skill splits the workflow into checkpoints, treats Excel as the source of truth, and derives reports from Excel to reduce inconsistencies caused by directly editing .md files.

How It Works

• Stage 0: Scans the MBU directory, generates a file list, and waits for user confirmation.
• Stages 1A/1B/1C: Reads the 01a, 01b, and 01c prompts to confirm table-document positions, extract data items, and identify business objects, producing {batch}_01_business_object.xlsx, JSON, and reports.
• Stage 2: Uses the stage-1 JSON as input to perform entity mapping, attribute design, relationship design, third normal form, CRUD, and data-standard work, producing stage-2 logical-model files.
• Stage 3: Runs five-dimension quality validation and produces stage-3 validation-result files.
• Stage 4: Optionally generates deliverables, including a multi-sheet workbook and a traceability matrix.
• Rebuild: Modifies only the relevant Excel file, then uses rebuild.md to regenerate downstream derived files from the affected phase.

Boundaries

Best suited for data-logic modeling workflows with an MBU directory, batch conventions, and phase prompt files. Human confirmation at checkpoints is still required; skipping validation is not recommended. Poor source documents can limit the quality of business objects, logical models, and validation conclusions.