Data Normalization

Overview

This skill defines 8 normalization patterns that can be applied to any ForemanOS Project Brain to fix common data quality issues. Each pattern is idempotent (safe to re-run) and produces a repair plan showing what will change before executing.

The patterns complement the detection layer (data-integrity-watchdog agent, conflict-detection-agent) by providing the fix side of the data health stack.

Prerequisites

Before applying any pattern, the AI must:

1. Read the target JSON file(s) in full
2. Read ${CLAUDE_PLUGIN_ROOT}/skills/project-data/SKILL.md for canonical key names
3. Generate a repair plan listing every proposed change (field path, before value, after value)
4. Present the repair plan to the user for confirmation
5. Apply changes via a single atomic Python script that reads → modifies → writes all affected files
6. Validate the output (re-read and confirm changes applied)

Repair Plan Format