Second-Brain Starter
Every repeated explanation is leverage leaking. Write stable context once, in a form the agent picks up automatically.
What’s in here
| File | What it holds |
|---|---|
CLAUDE.md |
The always-loaded context — stack, conventions, standards, how to work with you. The agent reads this every session. |
playbooks/ |
Step-by-step procedures for things you do repeatedly — deploys, releases, incident response. Reusable, not re-derived. |
memory/ |
Durable facts that accumulate over time — decisions, scars, what worked and what didn’t. |
The principle
There are three tiers of context, by how often it changes:
CLAUDE.md— slow-changing truth about the project and how you work. Loaded every time.playbooks/— procedures. Pulled in when the task matches.memory/— facts that accrue. One file per fact; an index so the agent can find them.
Do not put fast-changing task detail here; that belongs in the brief.
How to build yours
Build it by subtraction from repetition: when you explain something for the second time, write it into the right tier instead of explaining it again.
Adapt for your tools
This skeleton uses CLAUDE.md (Claude Code’s convention). The idea is tool-agnostic — Cursor rules, a CONVENTIONS.md, an AGENTS.md, whatever your agent reads automatically. The filename doesn’t matter. The discipline of writing context down once does.