What problem does it solve? Claude Code's default behavior is governed by two different mechanisms — instructions (CLAUDE.md) and enforcement (settings.json, hooks) — spread across four scopes (managed, user, project, local). Misplacing a rule means it gets ignored, overridden, or committed where it doesn't belong. This Skill turns a vague request like "stop asking me for permission" into the right edit, in the right file, at the right scope. ## Core Features & Use Cases - Request routing and scoping: Maps each request to the correct mechanism (instruction vs. enforcement) and scope (user, project, local, or managed/MDM fleet rollout). - Permission rule authoring: Writes allow/ask/deny rules, picks the right permission mode (default, acceptEdits, plan, auto, dontAsk), and escalates to PreToolUse hooks when string matching isn't enough. - Safe editing workflow: Inventories existing config, merges instead of clobbering, backs up files, validates JSON, and verifies changes with /status, /context, and /doctor. - Use Case: A team lead wants Claude Code to auto-run lint after every edit and block access to .env files for everyone on the repo. The Skill produces a committed .claude/settings.json with a PostToolUse hook and deny rules, warns about the workspace trust prompt, and explains how teammates add personal exceptions via settings.local.json. ## Quick Start Ask Claude to configure Claude Code so it stops prompting for your everyday build and test commands while blocking access to your .env files.
