Search skills/

Obsidian Second Brain Claude skill

Persistent, self-rewriting memory in your Obsidian vault: 48 commands, a shared cross-agent logbook (Crew), scheduled maintenance and a research toolkit.

eugeniughelbur/obsidian-second-brain 1 skill Memory & contextProductivity & notes
Our verdict Recommended with caveats

An unusually deep, tested memory system with a real cross-agent logbook, but it rewrites your notes automatically and its research commands depend on several paid APIs; read the freshness policy before trusting it with a real vault.

5/5 Depth expertise shipped Note Security Review before installing 1d Maintenance Updated today 5.0K Adoption stars OK License MIT, commercial OK

What Obsidian Second Brain does

This is one very large skill (about 90 KB) rather than a bundle: it turns an Obsidian vault into a knowledge base that Claude Code and other agents read, write and reconcile over time. The core idea, credited as an evolution of Karpathy's LLM Wiki pattern, is that new information rewrites existing notes instead of only appending, and contradictions are reconciled automatically under a documented freshness policy (OKM) that marks each stored fact as timeless, dated, or a pointer. A small CRITICAL_FACTS.md file (about 120 tokens) loads identity context at session start, with progressive token budgets for more.

Crew is a separate addition: a shared logbook, usable without Obsidian, so Claude Code, Codex, Cursor and Grok Bot can see what the other agents did and leave lessons that a human promotes to a rule with keep. The 48 commands include /obsidian-init, /obsidian-save, /obsidian-find, daily entries, kanban and people notes, decision and deal tracking, 4 scheduled maintenance agents, and a 7-command research toolkit (/research, /research-deep, /x-read, /x-pulse, /notebooklm, /youtube, /podcast) that calls Perplexity, xAI, Gemini, YouTube and OpenAI Whisper APIs when keys are set, falling back to free key-less sources (Wikipedia, HackerNews, arXiv, Reddit and others) when they are not. It ships 172 scripts, 111 tests and 10 eval files, needs Python via uv, and documents where its .env file lives on each OS.

Good for

  • People who already keep an Obsidian vault and want it to double as agent memory
  • Multi-agent setups where Claude Code, Codex and others should share a work log
  • Ongoing research workflows that save findings back into notes automatically

Not for

  • Users who do not want an agent silently rewriting their notes
  • Anyone unwilling to manage several optional third-party API keys

What's inside depth 5/5

One skill at 90 KB with 78 references, 172 scripts, 111 tests and 10 eval files in the repository, the top of the depth scale.

Skills
1
Reference docs
78
Scripts
172
Tests / evals
121
Templates
2
Avg SKILL.md
88.4 KB

Counted from the repository file tree. How we score depth.

Dependencies and costs

  • Obsidian vault (plain Markdown)
  • Python 3.10+ via uv for health check, research toolkit and bootstrap
  • Optional PERPLEXITY_API_KEY, XAI_API_KEY, GEMINI_API_KEY, YOUTUBE_API_KEY, OPENAI_API_KEY for the research commands
  • git
  • For Crew: a Grok Bot plan that includes it, and Codex's /hooks trust step
  • Telemetry: The research commands call third-party APIs (Perplexity, xAI, Gemini, YouTube, OpenAI) using keys you provide, each priced per call as documented in the README; no keys means the skill falls back to free, key-less sources. We found no telemetry sent to the project's own author.

Security review

Review before installing

An Obsidian-vault memory skill with commands, hooks and an MCP server that read and rewrite your notes. It registers SessionStart, PostToolUse and PostCompact hooks and edits ~/.claude/settings.json; an optional background agent runs headless with permissions skipped but is off until you set two environment flags. Several install and research scripts are missing from our copy, and the quick-install one-liner pipes the main branch to bash.

Worth knowing

  • Installs with curl | shRecommended install is curl -fsSL https://raw.githubusercontent.com/eugeniughelbur/obsidian-second-brain/main/scripts/quick-install.sh | bash from the unpinned main branch (also in SKILL.md line 75); quick-install.sh is not in our files.README.md:43
  • installs_hooksThe plugin documents a SessionStart hook, a PostToolUse hook on every Write/Edit/MultiEdit and a PostCompact hook; install.sh also adds the SessionStart hook to ~/.claude/settings.json. The PostCompact background agent ships inert until OBSIDIAN_VAULT_PATH and OBSIDIAN_BG_AGENT_ENABLED=1 are set; it then runs headless with permissions skipped but limited to Read/Write/Edit/Glob/Grep (hooks/obsidian-bg-agent.sh line 237; SKILL.md line 1236).hooks/hooks.json:2
  • Changes files outside your projectSymlinks slash commands and the skill into ~/.claude/commands and ~/.claude/skills, edits ~/.claude/settings.json, and (if you answer y) creates ~/.config/obsidian-second-brain/.env (mode 600) and opens it for your API keys; update.sh runs git pull in the checkout.install.sh:59
  • Installs packages or toolsIf you opt in to the research toolkit it runs uv sync in the repo to install Python dependencies.install.sh:116
  • Runs remote code at an unpinned versionThe bundled MCP server is started with uv run --no-project --with mcp<2 python .../server.py, resolving the mcp package at launch under only an upper-bound constraint..claude-plugin/plugin.json:15
  • Needs an API key for its own serviceOptional research commands read XAI_API_KEY, PERPLEXITY_API_KEY, GEMINI_API_KEY, YOUTUBE_API_KEY and OPENAI_API_KEY from ~/.config/obsidian-second-brain/.env and send each only to its own provider; the code that does so is not in our files.README.md:875
  • Makes network requestsResearch, podcast and YouTube commands call third-party APIs, and /notebooklm uploads the top matching vault notes to a Gemini File Search store when you run it.SKILL.md:1033
  • Relies on a CLI or package we could not reviewRelies on the uv, git and claude CLIs and on code under scripts/research that is not in our copy.hooks/obsidian-bg-agent.sh:237

What we checked

The repository (397 files) on Oct 29, 2026. Not a full review: 272 of 397 files were not downloaded (large repository; we reviewed the skill folders); we read 4 of 32 code files; it relies on code outside the repository: 272 repo files we do not have, including scripts/quick-install.sh, scripts/setup_settings_hook.py, scripts/platform-home.sh, the scripts/research toolkit, .env.example and integrations/obsidian-mcp-server/server.py; Not read in full: hooks/*.sh and *.py, adapters/lib.sh, and the Python scripts under scripts/ (inspected only by grep for network, subprocess, cron and delete operations); External services: xAI, Perplexity, Gemini, YouTube, OpenAI APIs.

NVIDIA SkillSpector static scan: 93 pattern hits (Agent Snooping 17, analysis-evasion 13, Privilege Escalation 10, Supply Chain 10, Rogue Agent 9, Excessive Agency 7). The YARA 'backdoor_persistence' hit is a sentence in architecture.md about not creating crontab artifacts; 'analysis-evasion' hits are the scanner noting partially analysed referenced files; credential hits are .env examples and the research key table. --dangerously-skip-permissions appears in documentation and in the opt-in background-agent hook described above.

A review, not a certification. Skills change after we check them — read the files before installing.

License

MIT — commercial use allowed.

Can it run unattended?

Yes. The README describes 4 scheduled agents that maintain the vault while you sleep and a Crew logbook with no background loop of its own, so unattended maintenance is a stated design goal, though the research commands and rewrites still warrant a later review.

Alternatives to Obsidian Second Brain

Obsidian Skills

kepano/obsidian-skills

Six skills by Obsidian's Steph Ango (kepano) for Obsidian Markdown, Bases, JSON Canvas, the Obsidian CLI, Defuddle and Knap.

Recommended Check first Depth 2/5

obsidian-wiki

Ar9av/obsidian-wiki

Forty skills that let an agent build and maintain a linked Obsidian vault from your notes, code and past agent sessions, then answer questions with citations.

Recommended Check first Depth 3/5

memU

NevaMind-AI/memU

Memory layer that mines agent session logs into a shared markdown wiki and re-injects it across Codex, Claude Code, Cursor and others, locally or via memU Cloud.

With caveats Broad access Depth 4/5

memsearch

zilliztech/memsearch

Cross-agent persistent memory for Claude Code, Codex, OpenClaw, OpenCode and DeepSeek Harness: Markdown files as the source of truth, with a Milvus search index.

With caveats Broad access Depth 3/5

FAQ

What is Obsidian Second Brain?

It is an agent skill that operates an Obsidian vault as persistent, self-rewriting memory for Claude Code and other agents, with 48 commands for saving, searching and maintaining notes, plus a cross-agent shared logbook called Crew.

How do I install it?

The README's quick path is a curl | bash installer, after which you restart Claude Code and run /crew setup or /obsidian-init, pointing it at your vault or a plain folder. It also supports Codex, Gemini CLI, OpenCode and others.

Do I need API keys?

No for basic note operations. The 7 research commands use paid APIs (Perplexity, xAI, Gemini, YouTube, OpenAI) when keys are set, and 2 of them fall back to free, key-less sources like Wikipedia and arXiv when keys are absent.

Is it safe to let it rewrite my notes?

The author documents a freshness policy for how facts are marked and reconciled, but automatic rewriting is still a real behavior change to your vault. Review the policy and consider versioning your vault with git before relying on it for important notes.

Sources: GitHub repository (stars, license, file tree, last push Oct 28, 2026). Review based on the README and SKILL.md files as published.