agentark-ai/AgentArk ? reverse-engineered prompt

Build me a self hosted personal AI assistant that runs locally in Docker and keeps my data private.

I want it to work like a secure workspace where I can chat with the assistant, save preferences and memories, schedule follow ups, and let it watch things in the background like email or reminders. It should be able to connect to common services like calendar, email, Telegram, Slack, webhooks, and web APIs, and it should also support creating reusable tasks or little apps from prompts.

Security matters most, so anything risky should ask for approval first, keep an audit trail, and stay sandboxed from my host machine unless I explicitly allow access. It should remember what I correct, learn from repeated workflows, and get better over time without sending my data anywhere else. A simple local web interface would be great, and if you need to check current docs for any integrations or model setup, go ahead and look them up online.