cimisy@2.2.0

Minor Changes

• 8211e9d: cimisy scan --full now also detects static, non-repeating content — headings, rich-text paragraphs, images, and standalone links — grouped into sections at semantic HTML5 boundary tags (<section>, <header>, <footer>, etc.), with a fallback region per component when no boundary tag is present. cimisy import presents these alongside collection candidates in one combined picker; selecting one writes a real .yaml/.mdx entry via writeSingleton, splices a singleton({...}) or page({...}){ sections } into cimisy.config.ts, and rewrites the source JSX to read from reader.singletons.<key>/reader.pages.<pageKey>.<sectionKey> — the same git-branch-review trust model as collection imports (clean working tree or --allow-dirty, a dedicated cimisy/import-<timestamp> branch, no runtime RBAC involved).

  Detection is conservative by design, mirroring the existing array scanner: content mixed with a non-literal expression (Welcome to {siteName}, {t("key")}), conditionally rendered content ({cond && <X/>}, ternaries), and ESM-imported images (import hero from "./hero.png") are reported as detected-but-not-eligible rather than guessed at. Only a narrow tag/prop allowlist is ever read as content (headings, p/blockquote rich text, figcaption, span, img/Image, a/Link) — className, data-*, aria-*, event handlers, and every other prop are never touched.

  A component rendered from a single route becomes a page-scoped section(); the same component rendered from multiple routes (e.g. a shared Footer) becomes a top-level singleton() instead.

  The admin content tree now separates a page's static content from its collections into two labeled groups instead of one flat list.

  No breaking changes — cimisy scan/cimisy import without --full behave exactly as before.