# personal-os-skills packages eight Claude Code skills for working inside an Obsidian vault > This repository publishes a Claude Code plugin marketplace of skills aimed at Obsidian: syncing Granola meeting notes, analysing Wispr Flow voice dictations, managing Obsidian tasks through the TaskNotes API, importing NotebookLM notebooks as linked graphs, and recalling context from earlier sessions. Each skill installs from the Discover tab and brings its own files and scripts. **ArtemXTech/personal-os-skills** — Claude Code skills for Obsidian | Claude Code x Obsidian Lab, 5 weeks live, starts Nov 10 - Repository: https://github.com/ArtemXTech/personal-os-skills - Website: https://lab.artemzhutov.com/ - Stars: 542 · Forks: 90 - Language: Python - License: MIT - Published: 2026-10-22 · Updated: 2026-10-22 · Language: en - Canonical page: https://hysenlabs.com/projects/artemxtech-personal-os-skills ## A plugin marketplace, not a library Personal OS Skills is described as Claude Code skills for Obsidian workflows, and the repository's shape makes that clear. The top-level entries are .claude-plugin/, skills/, docs/, a CLAUDE.md file, a LICENSE and a README. The .claude-plugin directory is the piece that defines the delivery mechanism: this is a Claude Code plugin marketplace, and the skills inside it are the payload. That framing explains the installation instructions, which are five steps and all of them happen inside Claude Code: ```bash /plugin marketplace add ArtemXTech/personal-os-skills ``` You run the marketplace add command in Claude Code, then run /plugin, go to the Discover tab, find the skill, select Install for you (user scope), and restart Claude Code. The restart is required, which is the standard behaviour of a Claude Code plugin rather than something specific to this repository. So the audience is narrow and well defined: people who drive Claude Code as their primary interface and keep their working notes in an Obsidian vault. The value proposition is that the integration is already written, rather than that it is clever. A skill is a directory with instructions and often scripts, and a marketplace is a way to distribute a set of them. The primary language is Python, the licence is MIT, the default branch is main, and the last push was on 2026-10-19. There are no GitHub releases, so a skill is installed from the current state of the repository rather than from a pinned version. One line in the repository description is promotional and worth setting aside: Claude Code x Obsidian Lab, 5 weeks live, starts Nov 10, with a link to lab.artemzhutov.com. It is a course or cohort being promoted alongside the code, and the skills themselves do not depend on it. ## The eight skills, and what each one actually touches The table of available skills is the substance of the repository, and each entry is worth reading for what it reads and what it writes. The granola skill syncs Granola meeting notes to Obsidian, and its Resources column says local cache, no API needed, which is the only entry that says so. Everything else in this list reaches out to something. The wispr-flow skill analyses voice dictation data from Wispr Flow, with resources listed as stats, search, export, dashboard. Four capabilities rather than one, and the dashboard implies some local server or generated page. The tasknotes skill manages Obsidian tasks via the TaskNotes API, and it is the one entry whose Resources column points to a video and a blog post rather than to a setup file, titled AI-Powered Task Management in Obsidian with TaskNotes and Claude Code. The notebooklm skill turns expert podcasts into personalized protocols with cited experiments in your morning routine, with a video. It is the most product-shaped of the skills and the least obviously a utility. The notebooklm-import skill imports NotebookLM notebooks into Obsidian as linked knowledge graphs, also with a video. The word linked matters: the import is not a flat copy of notes but a graph structure inside the vault. The recall skill loads context from previous sessions, with temporal and topic search through something called QMD and graph visualisation, and it carries both a video and a setup link to docs/memory-skills-setup.md. The sync-claude-sessions skill exports Claude Code conversations to Obsidian markdown with auto-sync hooks, and it shares that same setup document. Two skills pointing at one setup guide is the first sign that there is a cluster here rather than a set of unrelated utilities. The chief-of-ai skill routes work to persistent specialist agents and verifies the returned result, with a setup link to its own README under skills/chief-of-ai/. This is the one that is about the agent runtime rather than about the vault, and the phrase verifies the returned result is the interesting part: it is about not taking a subagent's word for it. Every skill lives under skills/ except tasknotes, which sits under docs/, and that inconsistency is the kind of thing you notice only by reading the table. ## Two of these skills are about memory, and they share a setup guide The recall skill and the sync-claude-sessions skill are clearly a pair, and the shared setup link to docs/memory-skills-setup.md is the evidence. Recall is about loading context from previous sessions, and it offers temporal search, topic search through QMD, and graph visualisation. Sync is about getting conversations out of Claude Code and into Obsidian as markdown, with auto-sync hooks so it happens without being asked. Read together, the pair is a memory loop: sync gets your past sessions into the vault, recall searches that vault and puts the relevant part back into your context. The two mechanisms serve different purposes, and the difference matters. Auto-sync hooks means the export happens as a side effect of using Claude Code, so the archive fills up without you doing anything. That is convenient and it is also a write path into your vault that runs automatically, which is worth understanding before you install it rather than after. The topic search through QMD is the part that is least explained in the table. QMD appears as an acronym with no expansion anywhere in the README, and it is the difference between the recall skill being a grep over markdown and being something with its own query semantics. The setup guide is where that would be documented, and the README's own instruction is to point at it. The graph visualisation is the other undocumented piece. A graph over sessions implies that the sessions are related to each other in a structured way, not just that they share keywords, which is a much stronger claim than search and would explain why the skill needs an index rather than just reading files. For anyone evaluating this repository, the practical sequence is to read docs/memory-skills-setup.md before installing either skill, because both of them write into the vault and both of them depend on something the README does not describe. ## The API-backed skills, and where a credential is needed Four of the skills talk to a service outside your machine, and the difference in what each one needs is worth separating. The tasknotes skill works through the TaskNotes API, which means a running Obsidian plugin on one side and an agent calling it on the other. The resources column for that skill is a video and a blog post rather than a setup file, which suggests the configuration is described in prose there rather than in the repository. The two NotebookLM skills are different in kind. The notebooklm one turns podcasts into protocols, which is a generation task against NotebookLM's own interface, and notebooklm-import brings notebooks across into the vault as linked knowledge graphs, which is a read and a write against a Google service. The granola skill is the odd one out, because its Resources column says local cache, no API needed. Whatever Granola produces locally is enough for it, which makes it the least risky of the eight to try first. The wispr-flow skill sits in between. It analyses voice dictation data that Wispr Flow produces locally, so the analysis itself may be local, but the dashboard and export suggest it writes somewhere you then look at. The general point for an adopter is that the security question is not uniform across the set. Some of these skills read a local cache, some of them call a third-party API with your credentials, and one of them, chief-of-ai, is about running subagents rather than about notes at all. A skill that holds an API key for a service with access to your notes is a different proposition from one that reads a file, and the README does not make that distinction for you. The repository does provide videos and blog posts for most of the API-backed ones, which is a reasonable way to see the behaviour before granting access. The README's links section also points to a YouTube channel and a Discord, so the community around these skills is reachable. ## Where this set of skills is the wrong choice Three limits, and the first is about the coupling to one agent and one note app. The first is the double coupling. These are Claude Code skills, so they only exist inside Claude Code, and they target Obsidian, so their output format is a vault of markdown notes with Obsidian conventions. An Obsidian user who does not use Claude Code gets nothing from the repository, and a Claude Code user who keeps notes somewhere else, a database, a wiki, a directory of plain files, gets a set of skills that write markdown into a structure they do not use. The README does not acknowledge this, and the table's descriptions all assume both halves. The second is that there is no version pinning. There are no GitHub releases, and the installation path pulls from the repository as it currently stands, so a skill can change its behaviour or its scripts under you. For skills that hold API credentials and write into your notes, that is a supply chain consideration rather than a nitpick. The skills are installed from a marketplace you add by URL, and they run with your permissions inside an agent that can read and write your files. If that is a concern, the mitigation is to read the skill directory before installing it, and the repository does keep the skills in separate directories under skills/ so they can be read individually. The third is the quality of the individual skills, which the README itself signals with an inconsistency. The tasknotes entry points its Resources at a video and a blog post while every other entry points at either a setup file or a feature list, and the chief-of-ai entry links to a README inside its own directory. The skills directory structure suggests some skills are more developed than others, and the README gives you no way to tell which is which beyond that uneven Resources column. The alternative approach is to write the small number of scripts yourself. Most of what these skills do is read a file from one place and write a file to another, and for the granola, sync-claude-sessions and recall cases, that is a cron job or a hook plus some markdown templating. The difference in approach is that the skills give you a working implementation, an agent that knows how to drive it, and a maintained update path, at the cost of running inside an agent with your credentials. A team that wants one narrow export, the sessions to markdown case, might prefer a shell script they can read in full. A team that wants the recall skill to be genuinely useful, with a graph over sessions and a query interface, would be building that themselves for a long time. ## Conclusion Adopt personal-os-skills if you already run Claude Code and keep your notes in Obsidian, since the skills are packaged as a marketplace you install once and then invoke individually, and the heavier ones, the Wispr Flow analysis, the NotebookLM import and the session sync, come with documentation you can check before installing. Do not adopt it as a general note-taking system, because these are Claude Code skills rather than an Obsidian plugin, so they only run when you are driving Claude Code. Verify first that the individual skill you want is listed with its resources, since the table lists resources for each skill, that the skill has a companion video or blog post you can watch before installing, that the API-backed ones, the TaskNotes skill and the NotebookLM ones, need credentials you are willing to hand to an agent, and that the repository's own promotion, a five week live lab starting in November, is not something your use of the skills depends on. ## FAQ ### How do I install the Personal OS Skills for Obsidian? Add the marketplace by running /plugin marketplace add ArtemXTech/personal-os-skills in Claude Code, run /plugin, go to the Discover tab, find the skill, select Install for you (user scope), and restart Claude Code. ### What skills does personal-os-skills provide? Eight: granola for syncing Granola meeting notes, wispr-flow for analysing voice dictation data, tasknotes for managing Obsidian tasks via the TaskNotes API, notebooklm for turning podcasts into protocols, notebooklm-import for importing NotebookLM notebooks as linked knowledge graphs, recall for loading context from previous sessions, sync-claude-sessions for exporting conversations as markdown, and chief-of-ai for routing work to specialist agents. ### Which personal-os-skills work without an API? The granola skill, whose Resources column says local cache, no API needed. The other entries reach out to a service, with tasknotes going through the TaskNotes API and the two NotebookLM skills against Google's service. ### What does the recall skill do in personal-os-skills? It loads context from previous sessions, with temporal search, topic search through QMD and graph visualisation. The sync-claude-sessions skill is its counterpart, exporting Claude Code conversations to Obsidian markdown with auto-sync hooks, and both point at the same setup guide. ### What does the chief-of-ai skill do? It routes work to persistent specialist agents and verifies the returned result, with a setup document at skills/chief-of-ai/README.md. It is the only entry in the list concerned with the agent runtime rather than with the notes vault. ### Under what licence is personal-os-skills released? MIT, with a LICENSE file at the repository root. The primary language is Python and the default branch is main. ## Sources - [ArtemXTech/personal-os-skills on GitHub](https://github.com/ArtemXTech/personal-os-skills) - [Issues](https://github.com/ArtemXTech/personal-os-skills/issues) - [License: MIT](https://github.com/ArtemXTech/personal-os-skills/blob/main/LICENSE) - [Project website](https://lab.artemzhutov.com/) - [README](https://github.com/ArtemXTech/personal-os-skills/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/artemxtech-personal-os-skills