Searches multiple literature sources and synthesizes a ranked research landscape report.
npx skills add https://github.com/zjunlp/Mechanist --skill msearch-zjunlp
Please help me install this Agent Skill. Skill: msearch Source: https://github.com/zjunlp/Mechanist/tree/main/skills/msearch Command: npx skills add https://github.com/zjunlp/Mechanist --skill msearch-zjunlp
High-intent search queries and answers about installing and using this skill.
Run /msearch followed by your query. It searches Zotero, Obsidian, local PDFs, web search, the arXiv API, and the mechanic-db cloud service in one pass, then merges and de-duplicates results into a single ranked set.
The base set covers Zotero, Obsidian, local PDFs, web search, arXiv, and the mechanic-db cloud database. Optional extras add Semantic Scholar for published venue papers, DeepXiv for progressive retrieval, and Exa for broad web content.
Yes. If a Zotero or Obsidian MCP server is unavailable, that source is skipped silently and the remaining base sources still run. Web search and arXiv are always available, so the search never fails due to a missing integration.
Yes, by passing the arxiv download option set to true. The top-ranked arXiv papers are downloaded to the papers/ folder, up to a configurable maximum, with rate limiting and size verification. By default only metadata is fetched.
Each run writes two files under msearch/<query-slug>/: RESEARCH_LIT.md with the raw retrieval dump including verbatim abstracts, and LANDSCAPE.md with the structured paper table, narrative synthesis, and structural gaps.
Tx1207Guide researchers through structured ideation with 5W1H framing and Zotero integration.
inference-simScan project code, tests, and documentation to formulate testable hypotheses.
cursorDesigns commands and processing loops that converge to the same state after crashes and retries.
plurigridQuery local ~/.topos notes and academic MCPs in parallel for multi-source research.
