semantic-model-optimizer Skill

Closes the loop in the semantic-model lifecycle:

semantic-model-analyst  →  semantic-model-deployer  →  (users customize in Semantius)  →  semantic-model-optimizer  →  …

The .md file this skill produces is interchangeable with one produced by the analyst: same front-matter keys (including version), same eight-section structure (§1 Overview, §2 Entity summary, §3 Entities, §4 Relationship summary, §5 Enumerations, §6 Cross-model link suggestions, §7 Open questions, §8 Implementation notes), same Mermaid diagram conventions. The deployer can re-deploy it; the analyst can audit or extend it. That compatibility is the main reason this skill exists, without it, live customizations in Semantius drift silently away from the .md and no other skill in the cycle can catch up.

Division of responsibility

• This skill owns the workflow: picking the module, reading its state, discovering related tables in other modules, transforming live state into the markdown template, and (opt-in) suggesting optimizations.
• The use-semantius skill owns the execution: every read is a semantius call.
• This skill is read-only against Semantius. It never writes to the platform. Any fixes suggested in Stage 5 are applied to the .md file only, a re-deploy via the semantic-model-deployer skill is how changes make it back to Semantius.

Writing conventions (apply to every output this skill produces)

These rules apply to chat output, the regenerated semantic-model markdown file, audit reports, and anything else this skill writes for the user to read.