bright-interaction

Second Brain For Claude — Design skill for Claude Code

Design community

Karpathy-style wiki + graphify knowledge graph + Claude Code hooks.

How to install Second Brain For Claude

This entry records only its repository, not the path inside it, so there is no exact command to give. Open bright-interaction/second-brain-for-claude and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Second Brain For Claude does

Karpathy-style wiki + graphify knowledge graph + Claude Code hooks. Self-deploying: clone it, tell Claude to set it up, done.

Alternatives in Design

Canvas — Open, create, or update a visual canvas — add images, text, PDFs, wiki pages, and banana-generated assets to O 3.3k ★
Plan Artifact — Creates one polished, self-contained HTML artifact from a plan, PRD, roadmap, strategy, migration, rollout, re 652 ★
New Module — Create a new LaraDashboard module with proper structure, CSS prefixing, hooks, menu, and CLAUDE.md 407 ★

README

Second Brain for Claude

A Karpathy-inspired wiki + knowledge graph + hook setup that stops Claude Code from re-exploring your codebase every conversation.

What it does

Teaches your Claude Code instance to look things up the way an experienced engineer would:

Query a knowledge graph first (fast index, low token cost)
Read a hand-curated wiki second (context the graph can't infer)
Only touch source code when the first two didn't answer

Enforced by hooks so Claude can't skip the order even if it wants to.

Quick start

Clone this repo, open it in Claude Code, and say:

Read SETUP.md and set this up in my environment.

Claude will ask you a handful of questions (project name, what to call your wiki, where it lives, etc.) and wire everything up: the wiki directory, hook config, CLAUDE.md, graphify install, and a verification run.

Expect 5 to 10 minutes for the first-time setup, most of which is graphify building its initial graph.

What you get

A wiki (you name it, default suggestion is your project name + "-brain") with entities/, concepts/, maps/ pages Karpathy-style
graphify (open source knowledge graph over your code)
Hooks that remind Claude of the work order on every prompt, tool call, and git push, plus the Boil the Ocean standard so partial fixes don't ship
A CLAUDE.md in your repo that documents the workflow for future sessions (and future teammates)
Guardrails against directory drift and credential leaks (optional, can skip)

Optional add-ons (opt-in during setup)

Stop hook - warns at session end if you have uncommitted or unpushed work (catches the "I forgot to commit" failure mode)
/adr slash command - capture architectural decisions into your wiki with one command
Scheduled jobs - daily TLS cert sweep, nightly Go test-coverage report, daily Claude-spend rollup, weekly security + SEO drift checks. macOS launchd +