Obsidian Second Brain

A cross-platform skill that turns an Obsidian vault into an AI-first second brain — agents read, write, link, and reorganize vault pages through 44 commands that build to Claude Code, Codex CLI, Gemini CLI, OpenCode, Hermes, and Pi. "One codebase, six platforms." Source: eugeniughelbur/obsidian-second-brain, https://github.com/eugeniughelbur/obsidian-second-brain, 2026-06-30

The premise: a notes vault is only as valuable as how easily an agent can maintain it. Rather than bolt one assistant onto Obsidian, the project ships a platform-neutral source that compiles into the command or skill shape each harness expects, so the same capture/organize/rewrite commands survive runtime changes. The repo sits at 2,824 stars; GitHub's latest release is v0.11.1 - Pi Coding Agent from 2026-06-28. Source: eugeniughelbur/obsidian-second-brain, GitHub, 2026-06-30

It went wide via @tom_doerr's one-line framing — "Rewrites Obsidian vault pages using AI agents" — the same dev-tools account that surfaced Stop Slop and Lean Ctx. The saved image captured an older v0.8.0 README card: one brain, four CLIs, 34 commands, and the core "vault rewrites itself" positioning. Source: X/@tom_doerr, 2026-05-31

What it does

• Cross-platform by design — one source tree builds to six runtimes: Claude Code (CLAUDE.md plus slash commands), Codex CLI native Agent Skills, Gemini CLI, OpenCode, Hermes native skills, and Pi prompts/skills. Same vault rules, runtime-specific dispatch. Source: repo README, 2026-06-30
• 44 vault commands — capture, organize, link, summarize, and rewrite vault pages; non-Claude builds ship 43 because /obsidian-calendar depends on the Google Calendar MCP.
• Architect + Retriever direction — v0.10 added /obsidian-architect for maintained codebase architecture notes; v0.11 releases added retrieval/Pi packaging and broadened the harness footprint. Source: GitHub releases, 2026-06-30
• Markdown-native — operates directly on the vault's flat .md files, so it composes with anything that reads an Obsidian vault.

Version snapshot

Relation to Kevin's wiki

Kevin's kevin-wiki is the same shape from the other direction: a flat-markdown, Obsidian-readable knowledge base (Karpathy LLM Wiki Setup Guide) where the agent is the writer and Kevin is the reader. obsidian-second-brain is the off-the-shelf, multi-CLI version of that contract, and its cross-CLI stance mirrors Everything Claude Code - Agent Harness Performance System's cross-harness AGENTS.md approach.

The useful steal is the single cross-harness skill source packaging: Kevin already maintains one AGENTS.md read by every harness, but many vault-maintenance commands still begin life in one harness before being projected. For session-memory specifically, Claude-Mem remains the dedicated compression layer; obsidian-second-brain is about durable vault curation, not transcript memory.

Adoption note

GitHub repo with a real README and stars → genuine skill candidate, not a guess. Run the AGENTS.md three-gate chain (Brin → skill-auditor → reputation) before installing, per the skills protocol. Source: repo README, 2026-06-12

---

Timeline

• 2026-05-31 | @tom_doerr surfaced obsidian-second-brain ("Rewrites Obsidian vault pages using AI agents"). 680 likes, 1,023 bookmarks. Repo ~2,407 stars; cross-CLI (Claude Code, Codex, Gemini, OpenCode), ~40+ commands. Source: X/@tom_doerr, 2026-05-31
• 2026-06-12 | Page created from coding-agents bookmark absorption; verified against the GitHub repo. Source: eugeniughelbur/obsidian-second-brain, 2026-06-12
• 2026-06-30 | Deep bookmark artifact review refreshed the page from the saved screenshot and current primary sources: image showed v0.8.0/four-CLI positioning, while current repo/release state is v0.11.1, six platforms, 44 commands, 2,824 stars, and Pi/Hermes support. Source: X bookmark artifact 2061225218547269909; GitHub release v0.11.1, 2026-06-30