Delivery Checklist

Code Contract

• export default includes every fnPath the host may call
• Reading pipeline implemented: searchComic / getComicDetail / getReadSnapshot / fetchImageBytes
• If download is supported, getChapter is implemented
• Plugin info, directory layout, and build scripts are based on the example repository
• Prefer BreezeHtml for HTML parsing; avoid bundling full cheerio

Data Structures

• Search list data.items[] fields are complete (id / title / cover / metadata / raw)
• Detail page data.normal.comicInfo + data.normal.eps structure is complete
• Chapter list and chapter detail use unified fields: id / requestId / logicalKey / storageChapterId / name / order
• Chapter images data.chapter.pages[] provide id / name / path / url
• All extern passthrough logic is closed-loop
• HTML parse results handle null / missing nodes so minor page changes do not crash the whole flow

Business Flows

• Favorite / like flow (toggleFavorite / toggleLike) works
• Comment feed (if implemented) supports pagination and replies
• Filters (if implemented) correctly emit core / extern

Stability

• Network requests use unified timeouts and error mapping
• Auth expiry returns structured unauthorized errors
• fetchImageBytes returns Uint8Array directly, with header x-rquickjs-host-offload-binary-v1: 1
• All ImageItem.url values are valid non-empty placeholders (not empty or 404)
• Both debug mode and release mode pass end-to-end

Release

• Bundle and plugin UUID are aligned (uuid is stable; do not change casually)
• Updated version in buildPluginInfo() in src/get-info.ts, then ran pnpm build
• Update channel configured: npmName and/or updateUrl (unlisted plugins need at least one, or silent update and Sync will not work)
• If using GitHub Release: tag matches version, updateUrl points to .../releases/latest
• Completed at least one full real-device regression (home → search → detail → read)
• Changelog covers new capabilities, behavior changes, and known limits