grandamenium/cortextos — reverse-engineered prompt

Build me a tool called cortextOS that lets me run persistent AI agents all day and night and control them from Telegram or my phone. I want a boss style orchestrator, an analyst, and optional worker agents that can share tasks through a simple bus, survive restarts, recover from crashes, and keep working for long stretches without losing context. I should be able to message it in Telegram to ask what happened overnight, approve actions, and create scheduled jobs like checking my inbox every morning at 8.

Please also include a clean web dashboard where I can see tasks, approvals, experiments, analytics, and overall agent health. I want risky external actions gated behind human approval, and I want the agents to do overnight research style experiments and report back clearly. A simple CLI for install, init, add agent, enable, status, doctor, list agents, and dashboard would be great, with smooth onboarding for API keys, Telegram bot setup, and the process manager. Default to Claude, but make room for other runtimes too. Look up current docs online if you need to.