A snapshot of one person's AI coding-agent configuration on one Mac. It is published so the arrangement can be read, audited, and adapted. It is not a supported product, and nothing here carries a guarantee that it is right for your machine.

No API keys, tokens, cookies, session state, or credentials of any kind. Secrets live in a separate secret manager and reach processes at run time. Configuration files that can hold a secret are templated to ${PLACEHOLDER} before capture, and a secret scan runs on every regeneration.

Also excluded on purpose: the machine's installed-application and package inventory with versions, install identifiers, git identity, and any operational runbook that names live infrastructure.

Paths appear as __HOME__ or ~, and the installer renders them for whoever runs it.

Open an issue describing the file and line. Do not include the value itself in the issue. If it is a live credential, say so in the first line so it can be rotated before anything else happens.

The install script writes to your home directory, replaces agent configuration files, and installs launch agents. Read install.sh before running it. Every agent configured here runs without per-tool permission prompts, which is a deliberate trade-off for one person's workflow and may be the wrong trade-off for yours. The gating rules in shared/AGENTS.md and the shell guard in claude-code/guards/ are what make that survivable; do not adopt the permission settings without them.