[skill-gap] Proposal: `database-and-schema-design` — schema design, migrations, indexing discipline

I've been using agent-skills in a production project and hit what looks like the biggest remaining gap in the Build phase: there's `api-and-interface-design`, but nothing on schema design, data modeling, migration discipline, or indexing strategy. Almost every production app touches a database, and agents are notorious for writing sloppy migrations (irreversible changes, missing indexes on foreign keys, destructive column drops without expand/contract).

**Pre-flight checks done:**

- Catalog: no existing skill covers this. `api-and-interface-design` treats schema only as an input to API shape; `deprecation-and-migration` covers system/API deprecation strategy, not DB migration mechanics; `ci-cd-and-automation` runs migrations but doesn't teach writing them.
- Open PRs: no overlapping proposal.
- Issues: #378 asks to deepen the DB substance of `performance-optimization` — I'd draw the boundary as: modeling/schema/migrations in this skill, query tuning stays in performance-optimization, with cross-references instead of duplication.

**Proposed scope:** entity/relationship modeling and normalization trade-offs; migration discipline (reversible by default, expand/contract for destructive changes, rollback tested); indexing discipline (FKs, query-shape-driven, no speculative indexes); constraints as correctness tools.

**Verification gates:** migration runs both ways against realistic data; rollback tested; schema change reviewed against actual query patterns.

Happy to contribute the skill + eval cases as a PR if this direction sounds right.

No files, tests, or entry points are named in the issue. Start by checking how `api-and-interface-design` and `deprecation-and-migration` are documented and linked in the project catalog and where their eval cases are defined. Then draft a sibling `database-and-schema-design` skill plus evaluation checks for reversible migrations and index discipline using the same format, and update catalog references; it is done when the new skill is added consistently and visible through the normal listing path.

Written by the indexing model from the issue text.