Answer 8 questions and download a ready-to-drop CLAUDE.md plus .claude/settings.json for your project. Permission allowlist, conventions, prohibited actions, all baked in.
.claude/settings.json so Claude doesn't prompt every run.Claude Code reads two project-level files to learn your codebase: CLAUDE.md at the repo root for human-readable conventions, and .claude/settings.json for permission rules. The settings file controls which Bash commands Claude can run without prompting (the permissions.allow array) and which it must always block (permissions.deny). Without these, Claude prompts before every command, which kills productivity in long sessions.
The generator builds both files from your answers. CLAUDE.md includes your project type, test/build commands, formatter, deploy target, and explicit prohibitions. settings.json mirrors your selected Bash commands as allow rules and adds your prohibited actions to permissions.deny. Drop both into your repo root (creating .claude/ if needed), commit, and Claude will pick them up on the next session.
Need a more advanced setup? Add hooks to .claude/settings.json to run formatters automatically after edits, or split global rules into ~/.claude/CLAUDE.md for machine-wide preferences. The generator outputs minimal, valid JSON you can extend by hand.