Third Brain Skills: a knowledge base and verification for agents
Third Brain Skills
21 skills build an Obsidian-style knowledge base with provenance, daily loops and verification before any done claim
- 140 ★ on the repository
- Last commit
Minimal access
Why this level
- The skills write and update wiki notes on disk, and semantic changes stay under your approval
- The command centre starts a local server and accepts commands, so enable it per the docs and only deliberately
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/third-brain-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills write and update wiki notes on disk, and semantic changes stay under your approval
- The command centre starts a local server and accepts commands, so enable it per the docs and only deliberately
This is third-party code. Review the repository files before installing.
What it does
Third Brain Skills is 21 skills that turn scattered chats into a lasting knowledge system. The main host is Codex, with compatibility adapters for Claude Code, Gemini, Cursor and Windsurf. The `wiki-ingest` skill takes a source (a link, article or PDF text), stages it unchanged, drafts a concept page and plans navigation updates. The `verify-before-claim` skill stops the agent from claiming fixed without test output and an exit code, and `daily-okr` runs a daily loop: one insight, one wiki update, one action, one output and a daily score. The other skills manage token cost, bounded loops, static dependency graphs and multi-agent teams. Version V8.1 builds work as a transactional pipeline: stage, author, graph plan, govern, commit and an optional deliverable. It runs with `python -m tools.worker_flow.cli`, and archiving happens only after a verified commit. Task profiles choose the amount of control needed, and semantic wiki writes stay under human supervision.
Who it is for. Users of Codex and other agents who keep Obsidian notes, are tired of losing context between sessions and want the agent's claims verified.
Good fit when
- The agent forgets context between sessions and you need a wiki with immutable sources
- Before saying done the agent must show test output and an exit code
- You want to run the working day as a loop: insight, wiki update, action, output, score
- Several agents need a shared pipeline with task owners and token accounting
- Knowledge from the wiki should update skills and instructions through supervised steps
Not a fit when
- You need one simple note without a wiki structure: the set is built for a complex knowledge system
- You do not use Obsidian or a similar vault: links and blocks rely on its format
- You want autonomous writing of semantic changes: such writes stay under human supervision
- You need documentation in English only: part of the guides in the repository are in Chinese
Example request
Use wiki-ingest on this article: stage the source unchanged, propose a concept page and a navigation update plan, then show a verification summary.Limitations
Codex is the main host, and other agents get only compatibility adapters. Some V7 guides and pipeline manuals are in Chinese. The last commit in the repository was made on August 21, 2026.
Installation uses the `install.sh` script and requires you to name the target for adapters explicitly. The command centre runs locally with `python tools/agentic_os_server.py --state-root .agentic-os-state` and listens on localhost only. Semantic writes need manual review, and the agent's own checks deserve a second look.
How to disable. Delete the skill folders from your agent's skills directory and the `.agentic-os-state` directory if you ran the command centre. Remove the Cursor and Windsurf adapter files from the project by hand.
FAQ
How do I install Third Brain Skills?
Clone the repository and run `bash install.sh codex`. On Windows PowerShell use `.\install.ps1 codex`. Adapters for other agents need an explicit target.
What does verify-before-claim do?
The skill requires test output and an exit code before any words like fixed or done. Without that proof the agent should not announce a result.
What is wiki-ingest?
The skill stores a source as an immutable note, drafts a concept page and plans navigation updates. It returns a verification summary, and entities are created in a separate reviewed step.
Do I need Obsidian?
The knowledge structure rests on notes with source provenance and block references, as in Obsidian. You can work with the wiki without the app itself, but the format follows its principles.
Is the command centre safe?
Per the README the server listens on localhost only and requires T02 runtime receipts. Start it only after reading `docs/agentic-os-command-center.md`.
Discussion
No questions or reviews yet.
Related
SkillsFour skills for a personal Obsidian knowledge base: an agent compiles articles and notes into a linked wiki, answers from it and health-checks it
An Obsidian vault template with 18 skills, 72 commands and 6 subagents for Claude Code: the agent turns sources into linked pages and answers from them
A skill set that lets an agent run a personal Obsidian wiki: it compiles knowledge from documents and past sessions and answers with citations
An agent saves web pages, PDFs and scans into a local vault, builds linked concepts and answers from them with source references.
cafe3310's personal skill collection: spec-driven project work, reviews, reports, browser, 3D and video in one repository
Four skills for a person with several AI assistants: a task register, evidence-based acceptance, beginner instructions
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Third Brain Skills: a knowledge base and verification for agents (this card) | no | 140 | ||
| obsidian-second-brain | no | 5,297 | a second-brain plugin for Obsidian without claim verification or agent pipelines | |
| LLM Wiki MCP | no | key is optional | 21,601 | an MCP server for LLM wikis without daily loops or the proof-first rule |
README in short
The README promises a lasting knowledge system for an agent and lists use cases. Then it shows a quick start, an example request, a before and after table, the V8.1 operating model with seven hard defaults and links to guides. It stresses separately that semantic writes stay supervised and automation is bounded.