argusappsec/argus ? reverse-engineered prompt

Build me an open source security assistant for codebases that feels more like talking to an analyst than reading a giant scanner report.

I want a single service that can review code with real security scanners, understand my company’s context, and explain what actually matters in plain language. It should work in a terminal chat, on GitHub pull requests, and through MCP so I can ask questions from my own AI tools. I also want a setup flow where I can give it my org’s profile, stack, and risk tolerance, then have that context carried into every review.

Please include a few built in review modes for things like quick PR checks, secret hunting, authorization issues, and threat modeling. Make it secure by default, keep a clear audit trail, and make sure it can answer follow up questions without giving the model unrestricted shell access. If you need to look up current docs or APIs while building it, go ahead.