Kritt-ai/open-kritt ? reverse-engineered prompt

Build me a self hosted web app that helps me find real vulnerabilities in code by running several AI agents on focused tasks instead of asking one model to scan everything at once.

I want a simple dashboard where I can create and reuse research workflows, point it at a local or remote repository, run a scan, and then see the findings grouped, ranked, and deduplicated in a clear way. It should also support validation steps after the scan, like running follow up scripts to check issues, build proof of concepts, and generate a report.

Let me connect my own model access, like Codex, Claude, OpenAI, Anthropic, or OpenRouter, and make the whole thing easy to run locally with Docker. Keep it private by default, with the frontend showing scan progress and results in a clean way. If you need to, look up current docs online for any provider setup details.