raye-deng/open-code-review ? reverse-engineered prompt

Build me an open source code review tool that checks AI generated code for the kinds of mistakes traditional linters miss, like fake imports, stale APIs, weird over engineered code, and obvious security problems.

I want it to work as a command line tool, a GitHub Action, and an MCP server so it can plug into tools like Cursor and Claude Code. It should scan a project, give a clear quality score, flag problems by severity, and output reports in a few useful formats like terminal, JSON, SARIF, and HTML. It should support local AI analysis with Ollama or any OpenAI compatible provider, and it should be able to run fully locally for privacy.

Please make it feel production ready, with good defaults, sensible docs, and a simple way to run a fast scan or a deeper scan. If you need to look up current docs online for any integrations, go ahead and do that.