vault-mind

Use when the user wants to check or clean up the vault's consistency/hygiene, e.g. "lint my vault", "find broken links", "any duplicate or empty notes", "check my frontmatter", "audit my tags", "update my MOCs", "sync my maps of content". Reports structural/metadata problems (broken links, dupes, empty notes, frontmatter drift, tag-taxonomy drift, out-of-date MOCs) and fixes the safe ones with approval. Never writes note prose.

Use when the user wants their notes on objective subject matter (math, physics, CS, philosophy, etc.) checked for things that are actually wrong, e.g. "audit my MSM notes for errors", "fact-check my philosophy of mind notes", "is anything in my

Use to set up (or re-profile) a vault for the vault-mind skills, e.g. "set up vault-mind", "init my vault", "vault-mind isn't finding my notes", "re-detect my vault structure". Inspects the user's Markdown vault, infers its layout and frontmatter schema, confirms with the user, and writes a vault profile that every other vault-mind skill reads. Run this once before using quiz/feynman/ask/etc. Writes only the profile file, never touches note content.

Use when the user wants to find concepts they should write notes for but haven't, e.g. "what notes am I missing on reinforcement learning", "find gaps in my philosophy notes", "what should I write next". Surfaces terms/concepts referenced across their notes and readings that have no note yet, ranked as a next-write queue. Never writes the notes for them.

Use when the user asks a question about the contents of their own vault, e.g. "what did I conclude about functionalism", "what do my notes say about lag time in MSMs", "which books did I rate above 8", "summarize what I know about Koopman operators". Answers strictly from their notes/readings, cited to the source files. Read-only, never writes.

Use when the user wants to find non-obvious connections between their existing notes, e.g. "what connects to my Koopman operator note", "find links I'm missing in my philosophy of mind notes", "connect my ML and chemistry notes". Surfaces conceptual overlaps that aren't wikilinked yet and suggests where THEY might draw the link. May insert a wikilink they approve; never writes note prose.

Use when the user wants to talk through something they just read (or are reading) to form their own view before writing a note about it, e.g. "let's discuss What Is It Like to Be a Bat", "I just finished Hyperion, discuss", "help me think through the Chinese Room argument". A Socratic partner that asks probing questions and challenges them; it does NOT hand them a take or write the note. This is for open-ended discussion of a reading to form a view; if they instead want to drill the formal structure of a single argument (reconstruct its premises, attack and defend it), that's /vault-mind:socratic-philosophy, not this.

Use when the user wants the actual source document for a reading pulled down locally so the agent can read it directly, e.g. "fetch the source for Epiphenomenal Qualia", "download the papers in my philosophy of mind list", "grab a local copy of the VAMPnets paper", "fetch everything I haven't downloaded yet". Resolves the reading's online-source URL, downloads a publicly available copy into the assets dir, and verifies it. Downloads source files only, never writes note prose.

Use when the user wants to explain a concept back in their own words to test their understanding (the Feynman technique), e.g. "let me explain backprop to you", "feynman me on the Koopman operator", "check my understanding of this concept". They explain; you find the gaps, errors, and hand-waving versus their note and the literature, and push them to sharpen it. Never writes notes.

Use when the user wants to test themselves / be quizzed / do active recall on a topic or note in their vault, e.g. "quiz me on Markov state models", "test me on this topic", "quiz me on stuff I haven't reviewed lately". Generates active-recall questions FROM their own notes, asks them interactively (interleaved, with a confidence check), grades answers against the note, classifies each gap, and flags misconceptions. Stamps a review-date for spacing but never writes note prose.

Use when the user asks for reading recommendations (books, essays, papers, videos) on a topic, e.g. "recommend readings on philosophy of mind", "what should I read next about X", "find me something on Y". Profiles their vault (what they've read, rated, and taken notes on), finds gaps and connections, then suggests specific, current readings grounded in web search.

Use when the user wants to actively learn a single philosophical argument by building and breaking it themselves, reconstructing it from memory, finding the load-bearing premise, attacking it, and defending it, the way you learn chess by replaying positions, not reading about them. Triggers e.g. "tutor me through Descartes' causal proof", "make me reconstruct the ontological argument", "I'm reading the Meditations and not digesting it", "drill me on Hume on induction", or just pasting an argument and asking to really understand it. This is the argument GYM, distinct from /vault-mind:discuss (open Socratic chat about a whole reading) and /vault-mind:feynman (explain a concept you already noted back). Do NOT use it when they just want a fact or a one-shot explanation and haven't signalled interest in working through it themselves.