Learning-first skills for a Markdown knowledge vault (Obsidian, Foam, Zettelkasten). The kit reads, quizzes, connects, and pressure-tests your understanding, but you write your own notes. The AI never writes note content for you.
/plugin marketplace add JEFF7712/rupans-ai-plugins
/plugin install vault-mind@rupans-ai-plugins
Then run /vault-mind:init once in your vault. It detects your layout and frontmatter schema
and writes .vault-mind/profile.md, which every other skill reads (that profile is what lets
the kit work on any vault shape). Local dev: claude --plugin-dir /path/to/vault-mind.
Assumes an atomic-note vault with YAML frontmatter and [[wikilinks]]; sparser vaults degrade
gracefully.
| Skill | What it does |
|---|---|
init |
Profile your vault for the other skills. |
ask |
Answer questions about your vault, cited to your notes. Read-only. |
recommend |
Reading suggestions from your taste and gaps, plus web search. |
fetch |
Download the public source behind a reading's URL. |
quiz |
Active recall from your notes: interleaved, confidence-checked, gap-classified, spaced. |
feynman |
You explain a concept back; it finds the holes. |
discuss |
Socratic partner on a reading, before you write the note. |
socratic-philosophy |
Drill one argument: reconstruct, attack, defend, consolidate. |
connect |
Surface missing wikilinks between notes (you approve each). |
gaps |
Your next-write queue: dead links, recurring un-noted terms. |
lint |
Vault hygiene: broken links, dupes, empty notes, frontmatter drift. |
audit |
Correctness pass: fact-check note content, web-verified, errors only. |
Skills are namespaced /vault-mind:<skill>. Only metadata, or by opt-in your own words, ever
gets written; never AI-authored prose.
recommend -> read -> discuss / socratic-philosophy -> you write the note
-> feynman -> quiz -> connect -> gaps (what to write next)
MIT.