RevoltSecurities/RAI — reverse-engineered prompt

Build me a terminal based AI security assistant for professionals. I want it to feel like an operator, not just a chat window. It should help with recon, API and web testing, code security reviews, vulnerability research, exploit or proof of concept scripts, and final reports, all from one terminal workflow.

Before it does anything substantial, it should write a clear step by step plan that I can review, approve, reject, or edit. Any risky action, like shell commands, file changes, or network requests, should ask for my approval first. It should also be able to run several specialist agents at once for things like recon, research, coding, and static analysis, then combine their results into one ongoing engagement.

I also want it to remember what it learned across sessions in simple markdown memory files, support background runs, and keep long conversations going without losing context. If a needed tool does not exist, it should be able to create a small script or helper for the job. Look up current docs online if you need to.