From local-library
Use when citekeys (`@Name2023`, bare `Name2023`), paper/book/chapter references, phrasing like "what does X argue", "from my library", "cite this", or any request that could plausibly be grounded in the user's local research corpus — reaches for `mcp__local-library__*` tools before answering from training, and treats `@citekey` as a library handle rather than a topic label.
How this skill is triggered — by the user, by Claude, or both
Slash command
/local-library:using-local-library-mcpThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
When a citekey or library-adjacent cue appears, the local library is the authoritative source. Reach for MCP tools before answering from training — the specific document in the user's corpus supersedes general knowledge of the author or work.
When a citekey or library-adjacent cue appears, the local library is the authoritative source. Reach for MCP tools before answering from training — the specific document in the user's corpus supersedes general knowledge of the author or work.
@Name2023, Name2023, Name et al. 2023| Intent | Tool |
|---|---|
| Get metadata for a known citekey | show_document @Name2023 |
| Search by topic across the corpus | search_library "topic phrase" |
| Browse or filter the corpus | list_documents (supports --status, --year, --author-contains, etc.) |
| Read a document's extracted text | get_document_text @Name2023 (note: >50 chunks triggers preview mode; follow up with range calls) |
@Name2023 (or unmarked but plausible citekey).show_document @Name2023 first.Error: No document found matching..., use the Did you mean @...? suggestion.⚠ USER-FACING ERROR:, surface the infrastructure problem — do not silently fall back to training.get_document_text @Name2023 or targeted search_library calls before writing.search_library: Starts with ### Results (N chunks from M documents); each result shows @citekey | score: X | *Title*, Section:, and Chunk index labels for follow-up.show_document: Markdown fields: **Title:**, **Authors:**, **Date:**, **Status:**, **Embedding status:**, **Chunks:**, **Original path:** (filesystem path for PDF fallback).get_document_text: Short doc (≤50 chunks) = full text in code block; long doc without range = preview + ### Sections outline; long doc with range = Chunks S–E of N label.| Excuse | Reality |
|---|---|
| "I know this author from training; I can just summarize" | The library has their specific document. One show_document call is cheaper than being wrong about what this version says. |
| "The citekey is just a label; the real question is topical" | If the user wrote a citekey, they mean that document, not a general topic. |
"search_library returned nothing, so the library doesn't have it" | Try show_document @<citekey> directly. Search can miss via embedding or exact-match issues; metadata lookup is authoritative. |
| "Listing all documents would be tedious" | list_documents supports filters (year, author, title, citekey prefix). Use them. |
| "I see the citekey but I recognize the work — I'll answer from memory" | Training knowledge is a guess. The document is on disk and one call away. Answer from the disk. |
| "The user probably knows their own paper — I don't need to verify" | Do verify. The user asked you to engage with their corpus, not with your assumptions about it. |
@Name2023 without invoking a library toolget_document_textError: No document found... as a reason to proceed from training instead of offering the fuzzy-match suggestion⚠ USER-FACING ERROR: as a signal to silently fall back to memoryGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub strophios/local-library --plugin local-library