Build a linked research wiki from source files
They used Cursor to turn papers and articles into a browsable markdown wiki with summaries, links, and an index inside Obsidian.
- What they did
- They opened the same vault folder in Cursor and Obsidian, put source files in a read-only raw/ folder, and used an AGENTS.md file to define how the wiki should be organized. Each time they gave the agent one source, it wrote a summary page, updated the index and related pages, and added an entry to a log. They read the summaries, guided the next steps, and later ran checks for contradictions and pages with no links.
- How it went
- The setup was described as a pattern, not a finished study, but it showed that one source could affect about 10โ15 pages. It also said the index-based approach held up for around 100 sources before a separate search tool became necessary.
- Tips
- Process sources one at a time and review what the agent writes before moving on. If the wiki grows past what the index can handle, add a search tool such as qmd.
Try it yourself with Cursor
You are helping me build a personal research wiki from source files. Read the materials in [source folder] about [topic] and create a markdown wiki in [wiki folder]. Make one page for each [concept, paper, or person] you identify, with a short summary and links to related pages, plus an index page that helps me browse the collection. Keep [raw folder] unchanged except for reading from it, and follow the rules in [AGENTS.md]. After each source, show me the pages you would create or update and pause for my approval before changing anything that would overwrite or remove existing content. Do not send, publish, delete, or discard any file without asking first. Continue until every source has been covered and the links are working.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Test Cursor on a real data science task
A data scientist graded Cursor on a Python football project and found it strongest for quick prototypes, but weak on pandas, cleaning, and scraping.
Let Cursor audit and summarize an Obsidian vault
Beatriz used Cursor to inspect her Obsidian vault, flag structure issues, answer note questions, and log what other sessions changed.
Draft an academic paper in Markdown with Cursor
A writer used Cursor to help draft and revise papers in Markdown while pulling citations from Zotero and a BibTeX library.
Set up a persistent study assistant
A student gave one assistant a name and used it for a week as a standing helper for research and everyday tasks.
Gather recent LLM papers and turn them into a Word file
Nitika Sharma used Perplexity Comet to collect recent LLM and agent training papers, then had it format the results in a Word document.
Turn genealogy notes into a spreadsheet and plan
Have an agent sort family-history notes into a clean spreadsheet and a short next-step research plan.