InfraNodus Skills

Knowledge graph + gap finder for your notes, vault or agent memory: shows what's missing and what to write next.

Give it a folder of notes, an Obsidian vault, an LLM wiki, a set of articles or a single text. It maps the main topics, shows how they connect, and turns the gaps between them into concrete prompts: what to research, link or write next. The graph comes from InfraNodus text network analysis, so it is measured, not guessed: word co-occurrence and network metrics, with the default extraction of a folder running locally without an LLM. That is why the gaps can be checked.

Made for AI researchers, Obsidian and second-brain users, researchers and content publishers. You don't need a terminal to start.

Get started without a terminal (Claude web / desktop)

1. Connect InfraNodus. In Claude, go to Settings → Connectors, add https://mcp.infranodus.com as a connector and sign in with your InfraNodus account (OAuth). Details: infranodus.com/mcp.
2. Add the core skill. Download infranodus.zip (latest release). In Claude, open Settings → Capabilities, turn on Code execution and file creation, then click Upload Skill in the Skills section and pick the zip. Other skills are uploaded the same way, one zip each (there is no all-in-one zip for Claude web); direct links are in Install a single skill.
3. Ask for the gaps. Paste a text, attach documents or give a URL: "Use InfraNodus to show the main topics in these notes and the gaps between them, then suggest three questions I should write about next."

Graphs of a whole folder, repo or vault (and keeping them up to date) need an agent that can read your files, such as Claude Code.

Claude Code quick start

Connect the InfraNodus connector first: in claude.ai, Settings → Connectors, add https://mcp.infranodus.com and sign in (Claude Code picks it up from your account), or run claude mcp add --transport http infranodus https://mcp.infranodus.com and sign in with /mcp. Then install:

/plugin marketplace add infranodus/skills
/plugin install infranodus@infranodus-skills

Open a project or vault folder and type /infranodus:infranodus. Want every skill in this repo? /plugin install infranodus-all@infranodus-skills. One skill at a time: see Install a single skill. Other routes (npx skills, install.sh, Cursor, OpenClaw, ChatGPT) are under Installation.

The workflow: gaps at the centre

Every stage feeds back into the gap analysis, and every gap suggests the next piece of research or writing.

research -> ANALYZE TOPICS & GAPS -> collect / curate -> refine -> publish
   ^                 |                                               |
   +-----------------+------------ next gap / next question <--------+

Stage	Skills	Example prompt
Research	infranodus	"Analyze this YouTube talk with InfraNodus and turn its gaps into research questions."
Analyze topics & gaps (centre)	infranodus, shifting-perspective	"/infranodus - graph this vault and show me the main topic clusters and the content gaps between them."
Collect / curate	llm-wiki, ontology-generator, project-learnings	"Set up an LLM wiki for my research on sleep and memory, with raw sources and wiki pages in Obsidian."
Refine	shifting-perspective, cognitive-variability	"My draft feels one-sided. What perspectives are missing, and how do I develop them?"
Publish	writing-assistant, seo-analysis	"Do keyword research for 'knowledge graphs for note-taking' and show what people search for that nobody covers well."

Skills in this repo

Core (start here)

• InfraNodus (infranodus, the core skill and default) - The front door. Turns a folder, Obsidian vault, LLM wiki or code repo into knowledge graphs and tells you what's missing: topic clusters, content gaps, research questions, central concepts, GraphRAG retrieval over the saved graphs. Invoked inside a project (/infranodus), it mines docs, PDF text, docstrings, WHY:/NOTE: comments and commit/PR/issue history deterministically (stdlib Python, no LLM), and keeps the graphs current: repo2statements.py --detect lists the files that changed since the last build, --update extracts only those, and the uploader replaces their statements in place with the server's delete_statements tool, so no graph is deleted by hand or duplicated. Also a reference for every InfraNodus MCP server tool (text analysis, comparison, SEO/GEO, structured memory) for Claude Code, Cursor, OpenClaw or any MCP client.

Collect and curate

• LLM Wiki (llm-wiki) - Based on Andrej Karpathy's LLM Wiki proposal this skill offers a guided workflow for designing and scaffolding a personal LLM-maintained wiki — a persistent, compounding knowledge base where the LLM incrementally builds interlinked markdown pages from raw sources. Instead of re-deriving knowledge on every query (like RAG), the wiki extracts, cross-references, and synthesizes knowledge once, then stays current as new sources arrive. Covers the full setup: raw sources, wiki pages, output layer, and a schema that co-evolves with the project. Use it when you want to build a "second brain", organize research, or set up an Obsidian + LLM knowledge base workflow.

• Ontology Generator (ontology-generator, folder skill-ontology-creator) - Build structured ontologies, taxonomies, and knowledge graphs by extracting entities, defining relationships, and mapping conceptual structures for any topic or for a collection of texts.

This is the beginning of the README. Read the rest at the original. Read the full README ↗