Timeln Second Brain Skill

We recently introduced the Timeln MCP—the transport layer that lets any compatible agent read and write your Timeln knowledge base. That was the plumbing. Today we are shipping what sits on top: the Timeln Second Brain skill, an open-source reasoning recipe for Claude Code, Cursor, and other agentic IDEs. On GitHub, one line to install.

What a skill is

A Claude "skill" is a folder with a SKILL.md file. Claude Code, Cursor, and other agentic IDEs that support the skill format auto-discover it, read the recipe inside, and trigger on matching phrases. No slash commands, no config — just say the words and the skill fires.

Our skill triggers on any of these: "second brain", "thinking partner", "knowledge partner", "what should I learn today?", "connect my ideas", "show my knowledge gaps", "based on my past data...", and a few more.

What it does

When triggered, the skill runs a structured pass over your real data:

1. Pulls context via the MCP—recent documents, topic subgraphs, and natural-language queries against your knowledge graph.
2. Applies MECE-style mapping—sorts entities into four quadrants: Known, Emerging, Isolated, Missing.
3. Applies PARA—classifies docs into Projects (ship), Areas (deepen), Resources (learn), Archive (stop).
4. Returns a pointed answer—not a search dump. Specific nodes, specific titles, one clear next step.

Example output from a real session:

## Your Brain, Right Now
You are building an agent layer over your own reading.

## MECE Map
             Connected                   Isolated
Known        MCP, LangChain, FastAPI     Pricing strategy
Emerging     Agent memory, RAG eval      Stripe webhooks

## PARA — What to Do Today
Project (ship):   Beehiiv integration brief — you started it Monday
Area (deepen):    MCP transport internals
Resource (learn): your 3 unread LangChain docs
Archive (stop):   that pricing thread from March

## The One Sentence
Your memory layer is real; your distribution layer is a guess — pick one and bet on it.

No fabrication. Every node, every title, every recommendation comes from a tool call. If the graph has nothing to say about a topic, the skill says so.

Why MECE and PARA

MECE (mutually exclusive, collectively exhaustive) is a consulting idea for decomposing a problem without double-counting or leaving gaps. On a knowledge graph, that maps to a simple test: a node is either well connected or not, inside a cluster or on its own. The isolated quadrant becomes your to-read list.

PARA is Tiago Forte's framework for sorting knowledge by actionability: Projects are time-bound and shipping, Areas are ongoing responsibilities, Resources are reference material, Archive is noise. Every document you save has a para_category; the skill uses it to turn "what do I know" into "what should I do."

Live visualization

Just type in the Claude Code or Cursor chat window — no scripts, no terminal:

> show my knowledge graph for MCP, LangChain and GTM

The skill queries your knowledge graph, merges the results, and writes a standalone kg_interactive.html file that opens automatically. Scroll to zoom, click to focus, search across nodes, filter by cluster. Gaps show up as dashed red edges so you can see what's missing instead of reading about it.

Install

npx skills add goodrahstar/timeln-secondbrain-skill

Works with Claude Code, Cursor, Zed, and any agentic IDE that supports skills. The command installs into the right folder automatically.

Then point your agent at the hosted MCP — see the MCP launch post for the JSON block. If a token is missing the skill returns a friendly "sign up at timeln.app" nudge; everything else stays zero-config.

Why open source

The thinking lives in the prompts and structure, not in a black box. We open-sourced SKILL.md so you can fork it, improve it, and port it to other agents. The durable advantage is the data—your knowledge graph, built over months of reading—not a hidden template.

Pull requests welcome.