eyeszik/agent-mission-control-os ? reverse-engineered prompt

Build me a web app called Agent Mission Control, for running an autonomous digital agency with a clear approval step before anything gets delivered.

I want a clean dashboard where a user can log in, create or view a tenant and project, submit a brief, and watch the workflow move through strategy, creative concepts, copywriting, design brief, campaign assembly, safety review, approval, and delivery. It should keep a full event history, let me replay what happened, and show whether a run is in a normal or degraded state if generation is missing. Please make the app feel like a real operations console, with analytics for runs, approvals, failures, and delivery outcomes.

Use the current stack already in the repo, with a Python backend and a Next.js frontend, and make local development work end to end. Keep auth and tenant access server side, include a safe approval gate before delivery, and do not allow live publication or paid media execution unless that is explicitly supported. If you need current docs for any library behavior, look them up online.