Recommendation Normalization Layer

The recommendation normalization layer is a deterministic validation pass applied before rendering optimized stack outputs. It tracks each tool's final disposition (KEEP, REPLACE, REMOVE, CONSOLIDATE, ADD), resolves contradictory states, removes duplicates, and prevents invalid target stack outcomes such as a tool being replaced and re-added later in the same scenario.

This layer sits between analysis output and UI presentation for StackScan recommendations. It ensures the optimized stack table, scenario metrics, and "why this changed" explanations all use one coherent recommendation state. Conflict resolution rules prioritize stronger actions (REPLACE/REMOVE) over weaker ones (KEEP/ADD), with lightweight dev warnings when conflicts are auto-resolved.

Recommendation integrity is critical for credibility. Users need to trust that suggested tool transitions are internally consistent and logically valid. The normalization layer protects against edge-case regressions and lets StackSwap present deterministic transformation guidance across AI, Balanced, and Cost strategies.