If you've been clicking "allow" on the same harmless bash commands every Claude Code session, this is for you. It analyzes your actual session history to find safe commands that keep triggering permission prompts, then offers to auto-allow them in your settings.json. The approach is evidence-based rather than prescriptive, so it won't suggest blanket rules or let you allowlist destructive commands like rm -rf. Works from any environment as long as your Claude config directory exists, though the changes only take effect in Claude Code itself. From the Every team behind Compound Engineering Plugin, which has picked up nearly 20k GitHub stars.
npx skills add https://github.com/everyinc/compound-engineering-plugin --skill claude-permissions-optimizer