技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
calibre-mcp【数据与研究】Search, read, and safely curate a Calibre ebook library through the calibre-mcp server. Use for finding books, metadata/full-text/semantic search, reading excerpts or chapter maps, listing figures and categories, auditing metadata quality, finding duplicates, building search indexes, or previewing gated library maintenance. Trigger on Calibre, ebook …
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Prefer the calibre_* MCP tools. This plugin connects to http://localhost:8080 and exposes write
tools because the user explicitly enabled them. Keep every mutation preview-first and narrowly scoped.
calibre_list_libraries; never assume names, IDs, or paths.calibre_ping when connectivity is uncertain.calibre_get_content with structure: true before walking a long book. Read bounded chunks
and preserve returned cursors only within the current extraction context.CALIBRE_MCP_ENABLE_WRITE=1 reflects explicit user authorization to expose write tools. Actual
modifications also require the Calibre Content Server to allow local write access.| Need | Tool/path |
|---|---|
| Connection check | calibre_ping |
| Find by title/author/tag/query | calibre_search, mode: meta |
| Exact words in books | calibre_search, mode: fts |
| Meaning-based discovery | calibre_semantic_search |
| Book metadata and formats | calibre_get_book |
| Outline or excerpts | calibre_get_content |
| Figures and illustrations | calibre_get_figures |
| Tags/authors/series schema | calibre_list_categories |
| Duplicate or quality audit | calibre_find_duplicates, calibre_quality_report |
| Metadata proposal | calibre_recover_metadata, calibre_extract_isbn |
| Search index | calibre_build_index |
For book hierarchy or TOC/spine analysis, use $calibre-analyze-book-structure. For Calibre
conversion selectors or EPUB package queries, use $calibre-write-xpath.
Use Calibre CLI only for file-level conversion/export or operations not exposed by MCP. If the GUI
is running, use the Content Server library URL for calibredb rather than --library-path.
ebook-convert input output is file-level and can operate on a copy without opening the library DB.