SPIKE: Propose Addressable Content Areas Spec

When an AI agent builds a diagram through the MCP server, it must reason about every coordinate and spatial relationship from scratch. This produces inconsistent layouts and wastes agent context on geometry instead of content.

Addressable content areas would let a human pre-define named regions in a draw.io diagram—think of them as slots or placeholders—that an agent can discover and write into by name. The agent focuses on what goes in each region; the diagram template controls where things go.

The exact form this takes is an open question. It might be container groups with metadata tags, `` wrappers with custom attributes, or something else that draw.io and mxGraph already support. This spike investigates the available primitives, evaluates implementation options with round-trip compatibility as a key constraint, and proposes a specification for the feature.