Remaining migration plan

This workshop no longer tracks the implemented type-system design. For the accepted design, see:

• ADR-012: types for structure, traits for review
• ADR-015: standardize authored type definitions on JSON Schema

What remains here is the deferred review-side follow-up.

Deferred work

1. Expand review sweep scope when the DB is stable

review_target_selector.py still discovers only top-level kb/notes/*.md files.

Recursive sweep expansion under kb/notes/** is deferred for now because broadening the review base while the DB and review tooling are still in motion would create unnecessary churn.

2. Verify parity after sweep expansion lands

Once recursive discovery is enabled, verify that:

• trait-gated gates are skipped for notes that do not carry the required trait
• direct note-local bundle runs and sweeps resolve the same applicable gates for the same note
• definitions and related-system reviews behave as expected once they are inside sweep scope

3. Revisit implied traits only if explicit traits become costly

The current system intentionally keeps trait computation simple: read traits: from frontmatter.

If explicit trait maintenance becomes burdensome, revisit whether some types should imply traits. Until then, keep the mechanism explicit.

4. Add more trait-gated review only when there is a real need

The current review integration is deliberately narrow. title-as-claim is the only trait with review-time routing effects today.

Future candidates:

• definition for term-precision or boundary-coverage review
• has-comparison for comparison-quality review
• has-implementation for implementation-specific review

Do this only when the review system reveals a concrete need.