← Back to Vibe Lab

Research & Working Skills

Tool OpenSource Ongoing

A collection of skills for research and everyday work. Selected skills are available on GitHub.

Project links
arXiv

Find papers by title, topic, exact keywords, or arXiv ID, then retrieve metadata and PDFs. Keyword searches use an explicit query-confirmation step; batch retrieval supports resuming interrupted work.

An arXiv ID retrieves a paper directly; titles and topics use fuzzy search; keyword searches require query confirmation before results, metadata, and PDFs.
View full-size diagram
China Travel MCP Installation

Prepare China-Travel-Planning-MCPs-All-in-One for an agent: check an existing installation, configure the chosen MCP client and local dependencies, then verify train, flight, map, and taxi capabilities.

Inspect an existing installation, reuse it if usable or configure the client, project dependencies and visible browser; build and verify the gateway before using train, flight, map and taxi tools.
View full-size diagram
Google Scholar

A coordinated workflow for paper search, advanced filters, cited-by exploration, pagination, full-text links, citation export, and troubleshooting. Browser and access checks precede the relevant operation.

Check browser access, resolve login or verification manually if needed, route the task to search, citation exploration, full text or export, and return structured results.
View full-size diagram
Literature Review

Turn a research question into a staged literature review: collect and deduplicate papers, screen titles and abstracts, optionally check the first two PDF pages, read selected papers in full, then synthesize a report and reading recommendations.

Define scope, collect and deduplicate, screen titles and abstracts, optionally check the first two PDF pages, read selected papers in full, then synthesize the review and reading recommendations.
View full-size diagram
PDF

A practical PDF workflow for creating documents, extracting text and tables, merging or editing pages, filling forms, and making scanned pages searchable with OCR. Validate the output against the original task.

Identify the task and document type, select creation, extraction, editing, forms or OCR, process the document and validate the output.
View full-size diagram
Xiaohongshu · Chrome DevTools

Search notes or open a supplied link, then extract visible text and metadata through the browser. Top-level comments are optional; login or verification prompts require user interaction before continuing.

Start from keywords or a note link, inspect access and visible content, search or open the note, optionally read top-level comments, then return structured content. Login and verification require user action.
View full-size diagram