# Claude Obsidian Skills: A Skill Library for Vault Automation

> Claude Obsidian Skills bundles reusable agent skills for Obsidian and Markdown workflows, from link fixing to video generation. It is a practical collection for Claude Code and Gemini CLI users, but its value depends on how well each skill fits your vault.

**jykim/claude-obsidian-skills** — Claude Skills for Obsidian Vault Management

- Repository: https://github.com/jykim/claude-obsidian-skills
- Stars: 49 · Forks: 10
- Language: Python
- License: MIT
- Published: 2026-08-23 · Updated: 2026-08-23 · Language: en
- Canonical page: https://hysenlabs.com/projects/jykim-claude-obsidian-skills

## What This Skill Pack Actually Solves

The README lists eleven skills, split into categories: PKM management, Obsidian-specific tasks, markdown, video, and image generation. The Obsidian skills are the core, covering canvas creation, link formatting, YAML frontmatter standardization, markdown structure validation, and Mermaid diagram creation. The video and image skills pull in external APIs and tools, which changes the setup burden considerably. For a user whose vault is plain text and who only wants link and structure fixes, the zero-dependency Obsidian skills are the relevant part. For someone hoping to produce narrated videos from slides, the skill pack expects FFmpeg and an OpenAI API key, so the promise of "automation" comes with real infrastructure requirements.

## How the Skills Work: SKILL.md as the Unit of Automation

Each skill is a folder containing a SKILL.md file, optionally with reference and examples subfolders. The SKILL.md file holds the skill's metadata, including name, description, and allowed tools, plus when to use it, core rules, step-by-step workflows, and quality checklists. This structure is what makes the pack portable: any assistant that reads SKILL.md can load the skill, not just Claude Code. The README states that each skill is self-contained and documented within its own SKILL.md, so copying that single file to another assistant's skill directory should be enough. The mechanism is file-based, not API-based. The Obsidian skills perform file operations only, which means they read and write markdown files directly. The Mermaid skill, for example, avoids common diagram errors by enforcing rules about markdown in labels and wide layouts, presumably by checking the diagram source before output. This design keeps the core skills lightweight and dependency-free, but it also means they have no awareness of Obsidian's internal state, like the graph view or plugin settings.

## Getting It Running: Commands and Config Keys

Installation is straightforward for Claude Code. You clone the repository, then copy the skill folders into your local skills directory. The README gives the command: `git clone https://github.com/jykim/claude-obsidian-skills.git`, followed by `cp -r claude-obsidian-skills/*-* ~/.claude/skills/`. Claude Code then auto-detects the skills. For other assistants, you copy the SKILL.md file from the desired skill folder. There is no package manager or build step. The advanced route uses git submodules so you can keep a vault in sync with the upstream repo. The commands are `git submodule add ...` and `git submodule update --remote`, with symlinks created for backward compatibility. When cloning a vault that includes submodules, you need `git clone --recursive`. The external dependencies are listed per skill. For markdown-video and video-cleaning, you need FFmpeg, an OpenAI API key, Python 3.7+, and Pillow. For gemini-image-skill, you need a Gemini API key, Python 3.7+, google-genai, and Pillow. The README shows `pip install Pillow openai google-genai` and exporting `OPENAI_API_KEY` and `GEMINI_API_KEY` as environment variables. The Obsidian and markdown skills need no dependencies beyond the assistant itself.

## The Obsidian Skills: Link Fixing, Frontmatter, and Structure

The obsidian-links skill formats, validates, and fixes wiki links, using proper filename, section, and folder conventions. That is a concrete, high-value task because Obsidian's wiki links break when files move or rename. The obsidian-yaml-frontmatter skill standardizes YAML properties with consistent formatting and naming, which is useful for vaults that rely on frontmatter for dataview queries or templates. The obsidian-markdown-structure skill validates heading hierarchy and document organization, so it can catch a note where a level-3 heading appears under a level-1 without a level-2 in between. The obsidian-mermaid skill creates diagrams that avoid common errors, specifically markdown in labels and wide layouts. These four skills share a common approach: they encode rules that a human would apply manually, but they let the assistant apply them at scale. The README gives example prompts like "Use the obsidian-links skill to check and fix broken wiki links in my vault," which suggests the skills are invoked by natural language rather than by a dedicated command. That reliance on the assistant's interpretation is a potential weak point: the skill definition guides the assistant, but the assistant still decides how strictly to follow it.

## The Video and Image Skills: Where the Dependencies Multiply

The markdown-video skill converts markdown slides to MP4 with text-to-speech narration, optionally using Deckset. The video-cleaning skill removes silent pauses using AI transcription and FFmpeg. The gemini-image-skill generates images via the Google Gemini API with multiple models and aspect ratios. These three skills are the most ambitious but also the most fragile. They require external services (OpenAI for TTS and transcription, Gemini for image generation) and local binaries (FFmpeg, Python). The README lists these dependencies but does not provide troubleshooting guidance for common failures, such as missing codecs, rate limits, or API quota exhaustion. The video pipeline involves multiple stages: slides to narration, narration to audio, audio to video, then possibly cleaning. Each stage can fail independently. For a user who just wants to fix links, these skills are overkill and add confusion. For a user who wants to produce video content from markdown, the skill pack offers a start, but the README does not explain how the skills handle edge cases like speaker notes, slide timing, or background music. The documentation is thin here, and the user must rely on the SKILL.md files themselves for details.

## Installation Models: Copy, Submodule, or Symlink

The README presents three ways to install: a simple copy, a git submodule for syncing, and symlinks for backward compatibility. The copy approach is the simplest but leaves you with a frozen snapshot that you must manually update when the repo changes. The submodule approach gives you a single source of truth and version pinning, but it adds Git complexity to your vault. The symlink approach is a middle ground, but it only works on filesystems that support symlinks, which is most Unix-like systems but not Windows without extra setup. The README explicitly mentions creating symlinks for skills like obsidian-canvas and markdown-slides, but it does not mention Windows. This is a real limitation for Windows users who want to follow the advanced path. The submodule update command, `git submodule update --remote`, pulls the latest commits from the default branch, which could introduce breaking changes if the skill definitions change. The README suggests pinning to specific commits if needed, but does not show how. The maintenance cost is real: if you use the submodule route, you must remember to run the update command and test your skills after each update, because the SKILL.md files may change their workflows or allowed tools.

## Limitations and Failure Modes

The most obvious limitation is that these skills are not Obsidian plugins. They operate on files, so they cannot interact with Obsidian's live graph, plugins, or settings. For example, obsidian-links can fix wiki link text, but it cannot update the backlinks pane or refresh the graph view; Obsidian will do that when it next indexes the file. The skills also depend on the assistant correctly interpreting the SKILL.md instructions. If the assistant ignores a rule, the output may be wrong, and the README does not describe any validation or rollback mechanism. Another failure mode is the external API dependencies. The video and image skills require API keys, and if the keys are missing or the quota is exhausted, the skill fails with an error that the user must debug. The README does not provide error handling guidance. The Mermaid skill avoids wide layouts, but the README does not say how it handles complex diagrams that inherently need width. The skills are also written for a specific set of conventions, such as "proper filename, section, and folder conventions" for links. If your vault uses a different convention, the skill may apply the wrong format. The README does not explain how to customize the skills beyond editing the SKILL.md files yourself.

## Alternatives: Obsidian Plugins and Other Skill Packs

The main alternative to using these skills is to rely on Obsidian's built-in features and community plugins. For wiki link fixing, Obsidian has a native "Update links on file move" setting, and plugins like "Obsidian Linter" can enforce formatting rules. Linter is a plugin that runs inside Obsidian and can fix YAML frontmatter, heading levels, and trailing spaces on save. It does not require an AI assistant or a SKILL.md file. The difference in approach is significant: Linter applies deterministic rules to every file in your vault, while these skills rely on an AI assistant to interpret and execute the rules on demand. Linter is more predictable but less flexible; it cannot understand context like "fix only the links in this folder" unless you configure it. Another alternative is to write your own scripts using the Obsidian URI or the Local REST API plugin, but that requires programming. The closest alternative in the AI-agent space is to use Claude Code's built-in file editing capabilities without a skill, but then you lose the structured workflow and quality checklist that the skill provides. The choice depends on whether you want deterministic automation (Linter) or conversational, context-aware automation (these skills).

## Maintenance, License, and Upgrade Path

The repository is licensed under MIT, which means you can use, modify, and redistribute the skills freely, including in commercial projects. The README does not mention any contribution guidelines beyond opening issues and submitting pull requests. The upgrade path is the git submodule update command, but the README does not describe how to handle breaking changes between versions. Since there are no releases listed, the project relies on the default branch, which means updates are continuous and may change behavior without notice. The maintenance cost for a user is low for the Obsidian skills, because they are just markdown files that you can edit. The cost is higher for the video and image skills, because they depend on external APIs that may change their interfaces. The README does not provide a changelog or migration guide. If you adopt the submodule approach, you can pin to a specific commit, but you must know which commit is stable. The project is part of the larger AI4PKM framework, which suggests it is actively developed, but the README does not state a release cadence. For a production vault, you should test each skill update on a copy before applying it to your main vault.

## Conclusion

Adopt Claude Obsidian Skills if you use Claude Code or Gemini CLI and want ready-made, file-based automation for Obsidian tasks like link validation, frontmatter standardization, and Mermaid diagram creation. Skip it if you prefer a single integrated plugin or need deep Obsidian API integration, since these skills operate on files and markdown, not the Obsidian app itself. Before adopting, verify that each skill's SKILL.md matches your vault's conventions, check the external dependencies for video and image skills, and test the skills on a copy of your vault to avoid unintended file changes.

## FAQ

### What do Claude skills actually do?

In this collection they are self-contained folders, each with its own description file, that you copy into an assistant's skills directory. The assistant detects them and selects one when a request names it, so there is no command syntax to learn. Skills range from pure file operations that need nothing installed to video and image skills that require external runtimes and API keys.

### What are Obsidian skills?

The five Obsidian specific ones: creating and managing canvas files with automatic layout, formatting and repairing wiki links against filename and heading conventions, standardising YAML frontmatter properties, validating document structure and heading hierarchy, and producing diagram code that renders in Obsidian while avoiding the usual label and layout mistakes. None of the five requires anything installed.

### What is Obsidian actually used for, in this repository?

The page assumes you already keep a vault and never defines the application. What it describes is the maintenance work around one: a knowledge base where links break, frontmatter drifts between notes, heading hierarchies flatten, and diagrams fail to render. The skills exist to keep those conventions consistent as the collection grows.

### Why do people use Obsidian with Claude?

The repository's answer is automation of the routine work: checking and repairing broken wiki links, standardising properties across notes, enforcing document structure, turning notes into knowledge maps and presentation decks, and narrating them into video. The pattern is that the assistant already reads and writes files, so the eight skills that need no external tooling can be adopted immediately.

## Sources

- [Official README](https://github.com/jykim/claude-obsidian-skills#readme)
- [Project repository](https://github.com/jykim/claude-obsidian-skills)
- [README](https://github.com/jykim/claude-obsidian-skills/blob/main/README.md)
- [Releases](https://github.com/jykim/claude-obsidian-skills/releases)

---

Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/jykim-claude-obsidian-skills