best-api

# Best API Own the public call-site decision. `VISION.md` owns product direction; `verdict` decides whether the direction earns work. The Plan playbook owns adoption and its review gates. Read public types, exports, runtime owners, representative consumers and docs. Inspect local dependency source and prior art before inventing an abstraction. Show the normal call and one real customization using actual public imports. Classify every noun by its honest package, registry or application owner. Compare keeping, deleting, merging, inlining and replacing. Keep a boundary only for an independent current job or a correctness, security, serialized data, native behavior or runtime law. Prefer one obvious common path with inference, useful JSDoc and progressive disclosure. Implementation difficulty does not establish good API design; published compatibility still needs an explicit adoption decision in this stable checkout. A runtime cost argument requires `benchmark`'s matched pre-acceptance probe. A type-only decision can record N/A with source evidence. A short call that hides the wrong owner is not an accepted design. For review, rank concrete findings P0 through P3 and give one recommended target with source, consumers, rejected alternatives and proof. A read-only review reports the repair and implements nothing. For an authorized repair, update the owning contract and every live example, doc, skill and check that teaches it. Search the old names and their word variants with `git grep --untracked`; distinguish release history from current teaching. Check Chinese twins, regenerate owned output and forward-test the real call. Keep unresolved public API teaching open in the plan.