Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This skill builds a Karpathy-style Obsidian wiki from articles, papers, and notes. Instead of re-retrieving raw documents on every query, it ingests sources and compiles them into a persistent web of concept, entity, and summary pages. Researchers, knowledge workers, and teams can then query the wiki, lint it for health, and file structured feedback that the agent applies as corrections. Knowledge compounds rather than being re-fetched, and human oversight stays durable across sessions.
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
A destination path to substitute for `<wiki-root>`.
A research topic to substitute for `<research-topic>`.
Python 3, because the documented scaffold command uses `python3`.
Using the llm-wiki skill, scaffold a new wiki at `<wiki-root>` for the research topic `<research-topic>`. Treat both angle-bracketed values as placeholders and replace them with my chosen destination path and topic.
A Markdown source copied into `<wiki-root>/raw/articles/`, with its filename substituted for `<source-file>.md`.
Using the llm-wiki skill, ingest `raw/articles/<source-file>.md` from the wiki at `<wiki-root>`. Replace `<source-file>` with the copied Markdown filename and `<wiki-root>` with the wiki's root path. Start by reading that wiki's `CLAUDE.md` and `wiki/index.md`.
A specific question to substitute for `<question>`.
Using the llm-wiki skill and the wiki at `<wiki-root>`, answer: `<question>`. Replace `<wiki-root>` with the populated wiki's root path and `<question>` with the specific research question. Base the answer on the wiki corpus and file the durable answer back into the wiki.
Claude
Codex
OpenClaw