Habib1001-m/hakim ? reverse-engineered prompt

Build me Hakim, a small toolkit for AI coding agents that helps them make safer, better justified changes without getting overly rigid.

I want it to guide the agent through a simple decision path before writing code, like checking whether something already exists, whether built in tools are enough, and only then using custom code if needed. It should also respect hard boundaries around safety, privacy, migrations, rollback safety, accessibility, data integrity, and user trust.

Please make the project work cleanly for the supported AI coding hosts, with the plugin and command experience wired up so a user can install it, start a session, and use the Hakim help and review style commands right away. Keep the installs and runtime behavior tied to the exact frozen version details in the repo, and make sure the documentation reflects the real current status honestly. If you need to look up current host docs online to confirm the integration details, do that.