A CommonMark wikilink extension: extract [[Target]] / [[Target|label]] from a
page body (skipping fenced + inline code, preserving offsets), and resolve each
target through the union — resolved is a link, unresolved is a createable
red-link (never a dropped reference). CamelCase auto-linking is off by default,
opt-in per space, and never double-counts a target already inside [[...]]. Link
model + resolution are core; rendering stays L6. New views/ package.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
