Obsidian Second Brain
Turns an Obsidian vault into a continuously maintained AI knowledge base.
The skill requires reading vault-local rules first, searching before creation, preserving sources, avoiding fabrication, and applying partial controls for calendars, sensitive content, raw sources, and logs. However, it authorizes broad reading, writing, cross-file propagation, self-rewriting, background maintenance, and external research; calendar scheduling and remote curl|bash installation create external effects, while confirmation, least privilege, isolation, and unified rollback are incomplete, so points are deducted.
The document gives clear access-method ordering, operating steps, naming conventions, anti-fabrication rules, and some atomic-write and logging principles, making the main path plausible. Static evidence does not include complete implementation coverage, abnormal-input handling, or end-to-end failure feedback, and behavior depends on multiple clients, MCP tools, scripts, and optional APIs, so the score is conservatively limited.
The audience and Obsidian use cases are clearly described, covering capture, search, research, daily notes, tasks, and calendars. Non-fit boundaries, trigger precision, client capability differences, Chinese-language support, and mainland-China network reachability are not sufficiently specified; some research functions depend on external services, so points are deducted.
SKILL.md provides a Quick Start, layered principles, command descriptions, naming rules, and frontmatter conventions, and the metadata states an MIT license. Installation dependencies, compatibility, troubleshooting, maintenance ownership, update paths, and limitations are not consolidated or fully disclosed, and the document depends heavily on unshown references and command files, so points are deducted.
The skill has broad coverage and specifies propagation into relevant notes, boards, logs, links, and sources, which could reduce repetitive knowledge-management work. Static review cannot confirm the actual completeness, output usability, or cross-platform behavior of all 45 commands; automatic propagation and rewriting may also increase review cost, so the score remains below the static ceiling.
The document includes a link to the Karpathy pattern, requires preserving source URLs, and describes some audit-log behavior, providing limited traceability. The available evidence is primarily author documentation and capability claims, without independent reproduction, third-party validation, or committed key-path test evidence in SKILL.md, so the score is low.
- Before first use, review the vault path, _CLAUDE.md rules, background agents, and calendar-write permissions; automatic propagation may modify many notes and structural files.
- Research and calendar features depend on external clients, MCP, API keys, or overseas services; verify mainland-China reachability, cost, data transfer, and key handling.
- Do not execute the documented curl|bash installer without review, and prepare independent backups and recovery procedures for bulk rewrites and unattended jobs.
What does this skill do, and when should you use it?
Obsidian Second Brain is a cross-CLI Agent Skill that connects an Obsidian Markdown vault with Claude Code, Codex, Gemini, OpenCode, Hermes, and Pi. It provides 44 commands for capturing conversations, managing projects and tasks, searching notes, importing research, reconciling contradictions, and generating syntheses. Its AI-first note rules prioritize provenance, recency, confidence, and cross-links. It is a strong fit for users who want durable, local Markdown knowledge management and are comfortable with a comparatively involved setup.
Reads and writes Markdown files in an Obsidian vault, checks _CLAUDE.md, templates, indexes, logs, boards, and existing notes, and runs Python, shell, and Git helpers. It supports keyword search and optional semantic search, and can call Perplexity, Grok, Gemini, YouTube, podcast RSS sources, its bundled Obsidian MCP server, and Google Calendar MCP. It creates or updates daily, people, project, task, research, architecture, synthesis, and audit notes while maintaining indexes and operation logs.
- An individual using Claude or Codex who wants meetings, decisions, tasks, and people captured into linked Obsidian notes.
- A developer who wants repository modules, dependencies, entry points, and key decisions maintained as architecture notes.
- A researcher who needs web pages, X posts, YouTube videos, podcasts, or PDFs converted into sourced notes with freshness markers.
- A project lead managing ongoing work through kanban boards, daily notes, recurring obligations, and reviews.
- A privacy-conscious Obsidian user who wants local Markdown ownership and optional local Ollama semantic search.
What are this skill's strengths and limitations?
- Stores data as ordinary Markdown and works with Obsidian Sync, iCloud, Syncthing, and Git-based syncing.
- Covers notes, tasks, projects, boards, research, code architecture, calendars, and synthesis in one workflow.
- Requires source URLs, freshness markers, historical facts, and cross-links for stronger provenance and retrieval.
- Core vault commands do not require API keys, and some research commands fall back to free sources.
- Provides build outputs for Claude Code, Codex, Gemini, OpenCode, Hermes, and Pi.
- It depends heavily on local filesystem access, shell commands, Python, and uv, making setup more involved.
- Some features require network access, API keys, or MCP integrations, including X, Gemini, YouTube, podcasts, and Google Calendar.
- The material mentions 134 CI tests but does not provide test results or independent verification here.
- Many workflows and defaults are Claude Code-oriented; other platforms use generated build artifacts.
- Automatic rewriting and propagation can affect multiple notes, so users must trust the model and review safeguards.
How do you install this skill?
For Claude Code, the recommended path is /plugin marketplace add eugeniughelbur/obsidian-second-brain followed by /plugin install obsidian-second-brain@obsidian-second-brain. The classic path is git clone https://github.com/eugeniughelbur/obsidian-second-brain ~/.claude/skills/obsidian-second-brain followed by bash ~/.claude/skills/obsidian-second-brain/install.sh. For Codex, Gemini, OpenCode, Hermes, or Pi, clone the repository, run bash scripts/build.sh --platform <platform>, and follow the generated dist/<platform>/INSTALL.md. The supplied material does not document a direct ChatGPT or bare-API installation.
How do you use this skill?
Set OBSIDIAN_VAULT_PATH to the vault and read the root _CLAUDE.md before the first operation. To create a new vault, the documented command is uv run python scripts/bootstrap_vault.py --path ~/my-vault --name "Your Name". Example triggers include /obsidian-save to preserve a conversation, /obsidian-find "keyword" to search, /obsidian-task "task description" to add work, /research "topic" to create a research dossier, and /obsidian-architect to document a codebase. Research, semantic search, YouTube visual analysis, and calendar workflows require additional configuration.
How does this skill compare with similar options?
The project presents itself as an extension of Karpathy's LLM Wiki pattern, replacing append-only pages with rewrites, contradiction handling, and automatic synthesis. Its README also compares it with Notion AI and Mem, emphasizing local Markdown storage, open source, and reduced vendor lock-in.