My Claude Code home — the shareable half of it.
This repository is a generated, sanitized mirror of a private ~/.claude.
A manifest in the private repo decides what ships; a deny scan over the rendered
tree fails the sync if anything private survives (client names, internal hosts,
personal contact details, absolute home paths, credential-shaped values).
Nothing here is hand-edited — edits happen in the private original.
| Path | What it is |
|---|---|
CLAUDE.md |
The always-loaded global context: working style, output conventions, hard safety rules |
skills/ |
Intent-triggered skills — git/PR flow, e2e, diagnosis, design, research, cleanup |
commands/ |
Slash commands, mostly thin entry points into the skills |
agents/ |
Subagent definitions |
docs/ |
Long-form references the lean CLAUDE.md points at |
rules/ |
paths:-scoped rules that load only when matching files are touched |
hooks/*-src/ |
Go sources for the guard and lint hooks (claude-guards, memorylint) |
Personal and team memory, session transcripts, anything naming a client or an internal host, and the skills that only work against private infrastructure (servers, self-hosted Sentry/Plane, private CLIs). Some documents therefore reference a skill or doc that isn't here — that's the scrub, not a bug.
Clone into ~/.claude (or copy the parts you want). The hook binaries build
with go build -o ~/.local/bin/<name> . inside each hooks/*-src/ directory;
settings.json shows how they're wired.