Vaultr ships with a set of built-in skills. Agents pick the right skill automatically based on what you ask for — you don't need to name a skill explicitly, just describe what you want in plain language (English or Chinese both work, since every skill triggers on both).
This guide covers what each built-in skill does and how to talk to it. For how skills are installed, managed, and customized, see the Skills section in the main README.
- Skills Guide
- Table of Contents
- vaultr-notes — everyday note operations
- vaultr-compile-note — compile a note into knowledge
- vaultr-index-knowledge — build the knowledge index
- vaultr-memory — extract personal memory
- vaultr-quote-extract — pull out key quotes
- vaultr-recap — associative and periodic review
- vaultr-writing — beat-by-beat article writing
The general-purpose router for day-to-day note work: creating, reading, editing, deleting, and searching regular notes; capturing short notes; and querying the knowledge base. Most conversational note requests land here first, and it hands off to a more specific skill (like vaultr-compile-note or vaultr-recap) when your request needs one.
Use it for: saving a new note, finding an existing note, quick-jotting a thought, or asking what you already know about a topic.
Example prompts
Save this as a note under /reading: "..."
Find my notes about pricing strategy
Jot down a quick note: remember to follow up with Alex about the partnership
What do I know about network effects?
Turns a single raw note (a clip, a journal entry, a transcript) into one or more knowledge units — durable wiki-style pages for the people, concepts, companies, or ideas the note discusses. Existing units are updated rather than duplicated; only entities with real staying power become units, so throwaway details stay in the source note.
Use it for: turning a web clip or long note into structured, reusable knowledge instead of leaving it buried in a single file.
Example prompts
Compile the note /clips/network-effects.md into knowledge
Extract knowledge from this podcast transcript and update the knowledge base
Run compile on /journal/2026-07-01.md
Scans the knowledge base for units that aren't in any domain index yet and files them in (creating a new domain index if needed). Safe to run repeatedly — it only ever adds missing entries.
Use it for: keeping the knowledge base's domain indexes (AI.md, Business.md, etc.) up to date after compiling new notes.
Example prompts
Rebuild the knowledge index
Refresh the domain indexes — I've compiled a bunch of new notes
Update the knowledge index to include anything I'm missing
Extracts personal memory from your short notes and knowledge base into six structured files under /_memory/ — identity, preferences, goals, beliefs, people, and current state. The first run scans the last 90 days; later runs scan incrementally. Memories that stop being reinforced fade out over time.
See the Personal Memory section in the main README for the full setup (including a daily scheduled Mate Bot trigger).
Use it for: giving every future conversation persistent context about who you are, without repeating your background each time.
Example prompts
Please update my personal memory. I'm Skoo, currently building an AI-native notes app.
Run memory extraction — also scan /journal/ this time.
Refresh my personal memory base.
Pulls high-value verbatim passages out of a note — like pull quotes in a magazine, not just a punchline sentence but enough surrounding context that the insight stands on its own — and saves each one as a short note with a backlink to the source.
Use it for: capturing the 3-8 best passages from an article or clip as standalone, re-discoverable short notes instead of re-reading the whole source later.
Example prompts
Extract the best quotes from /clips/deep-work.md
Highlight the important passages in this note
Pull out the key quotes from today's article and save them
Cross-checks content against your knowledge base and surfaces what's related or in tension with it. Two modes:
- Associative recap — check a single note, draft, or pasted passage right now, while the thought is fresh.
- Periodic review — recap everything created or updated over a recent window (e.g. the past week or month).
Both modes produce a conversational summary only; nothing is saved to a file.
Use it for: catching contradictions before they calcify, or getting a digest of what you've been thinking about lately.
Example prompts
Recap this draft against my knowledge base — anything related or conflicting?
Does this note conflict with anything I've written before?
Give me a review of everything I wrote this week.
Leads a choose-your-own-adventure writing process: you state the topic, the intended reader, and the stance; the skill scans the knowledge base for relevant units, proposes a palette of sources for your approval, then writes the article one beat (one paragraph-sized move) at a time, offering a few candidate next beats after each one. When the article is done, it can be compiled back into the knowledge base, closing the loop.
Use it for: writing something grounded in your own accumulated understanding, rather than starting from a blank page or copy-pasting raw notes together.
Example prompts
I want to write an article about why SEO is dying, for indie hackers, arguing that content moats matter more than ever. Save it to /essays/seo-is-dying.md.
Help me write a piece on remote work culture, aimed at first-time managers, taking the position that async-first is usually the wrong default.