claude-setup

Scores a Claude Code environment 0–100, detects your engineering profile, and generates a tailored CLAUDE.md.

ToolPublic

`/setup` checks six layers — CLAUDE.md, settings and hooks, MCP servers (including OAuth MCPs it detects from live session context, since those never show up in settings.json), skills, credentials, and memory — and grades the result S through D. First run detects an engineering profile from installed tools (DevOps/Platform, AI Builder, Full-stack, Researcher, Enterprise) and tailors its recommendations to it.

Every fix shows a diff before it applies, merges into existing config rather than overwriting it, and re-runs the layer scan afterward to confirm it actually worked. A 10-step onboarding mode walks a fresh machine through setup with a verification gate at each step.

Install

curl -fsSL https://raw.githubusercontent.com/jmenzies722/claude-setup/main/install.sh | bash

# then, in any Claude Code session:
/setup

Why it is built this way

Honest scoring by design — the README states it plainly: a 60 is better feedback than a fake 90.

GitHub