SecondBrain

A personal, domain-organized knowledge repo that learns from my sessions and tweets and lets me recall it later by meaning — not just keyword search.

Mature domains eventually get distilled into reusable Claude Code skills under skills/ — browsable and installable via skills.sh (see Skills).

Layout

Domains: aptos, ethereum, stellar-soroban, ai-agents, databases, clippers (content-clipping learnings; campaigns like yahoo/solana/ touchgrass tracked as tags), general (cross-cutting TypeScript / React / Node / SQL). Add a new one by creating domains/ /notes/.

Skills

The whole collection is browsable + installable via skills.sh using the single-repo model — all skills live in skills/, one repo, one manifest.

The playbook skills (aptos-move, soroban, make-clips) read their domain's notes/ folder as a living source, so each newly distilled session makes the skill richer without editing the SKILL.md.

Note format (the contract)

Every note carries YAML frontmatter so both the catalog and the embedder can read it:

Usage

make auto-uses .venv/ if present. (Homebrew's Python is externally managed, so deps live in a project venv rather than globally.) The catalog builder (build_index.py) is pure stdlib and runs with plain python3 too.

To add knowledge: drop a raw file in sessions/inbox/ (or tweets/inbox/), then in Claude Code invoke the **distill-session** skill to file it as a note, and run make index.