AI Project 2nd Brain

A Claude-powered project knowledge system for Loka engagements. Fork it, configure it in 2 minutes, and have a working project brain that Claude understands from day one.

Structured the way a TPM thinks — so whether you're an ML Lead, Tech Lead, or shadowing engineer, the brain helps you work like a TPM would: tracking stakeholders, organizing by SOW, running working sessions, and generating TLUs.

What's inside

project-vault/         ← your project brain (open this in Obsidian)
  inbox/               ← raw captures land here first
  sows/                ← one directory per SOW
    _template/         ← copy this to create a new SOW
  stakeholders/        ← client stakeholder profiles
  team/                ← Loka team member profiles
  roadmap/             ← challenges, ideas, features for this engagement
  notes/               ← project-level thinking and context
  archive/             ← processed inbox items worth keeping
  templates/           ← working session, meeting summary, TLU templates
project.config.example.yaml  ← config reference
project.config.yaml          ← your project config, gitignored — never committed

Setup

1. Fork and clone

Fork this repo on GitHub (one fork per project engagement), then clone your fork:

git clone https://github.com/<your-org>/<your-project>-ai-2ndbrain
cd <your-project>-ai-2ndbrain

2. Run /onboard

Start Claude Code from the repo root:

claude

Then run:

/onboard

Claude will ask you for each config value, write project.config.yaml, replace all {{placeholders}} in the vault, and create your first SOW directory. Takes about 2 minutes.

project.config.yaml is gitignored — your config is never committed.

3. Enter the vault and bootstrap

cd project-vault
claude

Then run:

/bootstrap

This pulls in the SOW document, all historical meeting notes from Google Drive, and recent Slack history — then generates a context snapshot, stakeholder stubs, and a tensions/gaps report between what the SOW commits to and what's actually happening in the project.

4. Fill in any gaps

Open notes/project-context.md — that's your ground truth. Fill in team profiles in team/, enrich stakeholder profiles in stakeholders/, and address any tensions flagged by bootstrap.

5. Commit and push

git add -A && git commit -m "init: configure project brain"
git push

6. Orient yourself anytime

From the repo root, run /2ndbrain — it reads your current state and tells you exactly what to do next.

How to use it

Not sure where you are? → run /2ndbrain from the repo root — it checks your setup state and tells you the next step

Capture anything → drop it in inbox/ and ask Claude to process it

Start a working session → say "I need to work on [topic]" — Claude creates a slug-based session under your SOW

Generate a TLU → say "generate this week's TLU" — Claude pulls from working sessions and meeting notes

Track a stakeholder → copy stakeholders/_example/, fill in profile.md

Log a meeting → ask Claude: "log this meeting summary" — saves to the right SOW

Track an idea or challenge → ask Claude: "save this as an idea" — goes to roadmap/

SOW structure

Each SOW lives at sows/<sow-name>/:

Directory	Purpose
<sow-name>-reference.md	Scope, key dates, deliverables for this SOW
work/	Working sessions — slug-based, open-ended
meeting-summaries/	Dated meeting notes
deliverables/	Final deliverable artifacts

To start a new SOW: copy sows/_template/ and rename it.

Requirements

• Claude Code — for AI-assisted project work
• A markdown editor — Obsidian recommended (any editor works)
• Google Drive and Slack connected in Claude Code (for /bootstrap)