Research Wiki — AI skill for Claude Code

research-wiki

Turn a stack of deep-research documents into a wiki you can browse, search, and question.

`research-wiki` is a [Claude Code](https://claude.com/claude-code) skill. You point it at long research reports. It reads them, builds a linked set of notes, finds where the reports agree and disagree, and makes one web page you can explore. You run it with `/research-wiki` commands.

---

What it is

You start with source documents: long, deep research reports, one per question. For example, "What is wu-wei?" or "What is life?". A stack of files like this is hard to use. You cannot see how the ideas connect. You cannot find every place two authors disagree.

research-wiki sits between you and those documents. It reads them and builds three things: a linked wiki, a browsable atlas, and a searchable copy of the source text. The model does the reading and writing. You stay the editor.

The source documents are never changed. The skill only reads them.

---

What it does

• It reads each source and pulls out the key concepts and thinkers. Each one becomes its own note.
• It links the notes to each other, so you can see what relates to what.
• It compares the sources. Where several agree, it writes a theme. Where they disagree, it writes a debate, with both sides quoted and cited.
• It builds an atlas: one web page with a map of the ideas, a search box, and a reading pane.
• It makes the source text searchable, so you can ask plain questions and get answers from the real documents, not from a summary.

---

What you need to run it

• Claude Code. research-wiki is a skill you run inside it.
• Your source documents. One folder per report, as HTML or markdown. A separate research pipeline makes these. research-wiki only reads them.
• Python with a few libraries: markdown, beautifulsoup4, networkx, PyYAML (and pytest to run the tests).
• Two optional helper skills. The wiki still works without them, with l