Anthropic shipped a read-only plugin that maps your dependency graph to specific automations.

Claude Code Setup reads package.json, language files, and directory structure and recommends automations across MCP servers, skills, hooks, subagents, and slash commands. The mappings are hardcoded decision tables rather than model judgment: React detection suggests Playwright MCP, Stripe or auth code suggests a security-reviewer subagent. Never modifies the codebase, so it's safe to point at a repo you don't own.