Search and retrieve documents from local markdown knowledge bases using QMD.
npx skills add https://github.com/Yassimba/loom --skill qmd-yassimba
Please help me install this Agent Skill. Skill: qmd Source: https://github.com/Yassimba/loom/tree/main/skills/qmd Command: npx skills add https://github.com/Yassimba/loom --skill qmd-yassimba
qmd query with intent, lex, vec, and hyde fields for conceptual recall. - Source Retrieval: Fetch full documents by docid or path with qmd get and qmd multi-get, including line-numbered output and line-range slicing via the :from:count suffix. - Collection Management: List, filter, and maintain indexed collections with qmd collection, qmd status, and qmd doctor diagnostics. - Use Case: A user asks what their notes say about staying data-informed without becoming metric-driven. The agent runs a structured qmd query against the concepts collection, retrieves the matching note with qmd get, and answers with cited docids and line numbers. ## Quick Start Use the qmd skill to search my notes for the concept about metrics as instruments and retrieve the full source document.High-intent search queries and answers about installing and using this skill.
Run qmd search with exact terms for BM25 lexical matching, or qmd query with structured intent, lex, vec, and hyde fields for conceptual recall. Then retrieve full documents with qmd get or qmd multi-get using the returned docids.
Use qmd search when you know exact words, titles, names, or rare phrases, since BM25 is faster for lexical matches. Use qmd query with structured fields when the user describes an idea indirectly or needs semantic recall across differently worded sources.
Append a :from:count suffix to the path or docid, such as qmd get "#abc123:120:40" to read 40 lines starting at line 120. Do not pipe output through sed, head, or tail, since that breaks docid resolution and line numbering.
Only model-backed commands like qmd query, vsearch, and reranking depend on local models or GPU setup. If those fail, run qmd doctor for diagnostics and fall back to qmd search with stronger lexical terms.
Yes, QMD exposes an MCP query tool that accepts structured searches with lex, vec, and hyde types plus an intent field and collection filters. The Skill supports both the qmd CLI and the MCP server interfaces.
Bare queries that paste the user's wording rely on the built-in expansion model and lose context. Write the intent field yourself to steer ranking away from nearby-but-wrong concepts, and add lex or vec terms with domain vocabulary.
moistalgiaSearch local markdown notes and documents with hybrid keyword, vector, and reranked retrieval.
real-caseSearch and synthesize project documentation into cited answers.
XSpoonAiIndex local notes with vector embeddings for offline semantic search.
indigoai-usSearch HQ knowledge and indexed codebases with qmd semantic search and Grep fallback.