C.3.3:8 - Authoring, Review & Rating Guidance (informative)

Compare identity before authoring a bridge. For a changed locality, source, team, spelling, or scheme, first repeat the kind-definition comparison under C.3.1. Stop without a bridge when the same kind continues.
Start narrow and honest. For two distinct kinds, declare only the directional correspondence and subkind facts the receiving use actually relies on; mark any required but unsettled order result unknown.
Prefer an existing suitable target kind. If the target already has a suitable kind and declaration edition, relate that kind directly. If a new target declaration is required, author it separately before asserting bridge obtaining; list what the mapping predicate preserves, relaxes, or drops.
Write loss notes in plain language. Example: “EV vs ICE subkinds collapsed; battery‑health invariants dropped.”
Fix the definedness area. Bind to target Standards/versions and any environment selectors essential to classification.
Assign CL^k from exemplars. Calibrate on concrete counter‑examples and preserved properties; resist optimistic ratings.

Identity checked first? Same kind reused without a bridge, or two distinct kinds and the obtaining correspondence shown? Add Scope or F.9 relations only when the receiving use consumes them.
Order claims honest? Any ⊑ inversions? Collapses disclosed?
CL^k plausible? Based on preserved properties, not name similarity?
Loss notes present? Will they force narrowing of Scope or extra tests?
Definedness area clear? Guard will fail closed outside it?
Penalties wired to R? No hidden tweaks to F/G?

High CL^k: signature equivalence or up‑to‑iso; ⊑ fragment preserved; only cosmetic losses.
Medium CL^k: some invariants relaxed or lost; selected subkinds collapsed; order preserved on critical path.
Low CL^k: name‑only correspondences; properties diverge; order not preserved. Establish that the KB-01 correspondence predicate holds before bridge use, even for a name-only match. Expect significant R penalty and/or adapters.