We are working on the Subordinate Agents Editor, that will allow you to override default prompts and give custom instructions/knowledge to specialized AI agents right away, from the UI, just like you do with Projects. Meanwhile, use the `a0/agents/` directory as your reference, and let's go through the creation of a new agent. Easy way: ask A0 to look into a0/agents and walk you through the creation of your new set of prompts for the new agent. But let's see how it works internally. For example, look at `a0/agents/hacker/prompts`: notice how it overrides the environment and the main role of the agent with `agent.system.main.environment.md` and `agent.system.main.role.md`. 1 - Open the A0 File Browser 2 - In `a0/prompts` you’ll find all the default framework prompts. 3 - These cover every step of the pipeline, including things like the summarization prompt used when a chat gets too long. All of these can be customized if needed. 4 - Take your time to review what you actually want to override or edit, for instance, the communication prompt if you want the agent to output custom fields after “Thoughts” or to change its thinking style, or again the role prompt to make it more robot-like "Executing task... Running code" in a concise way without conversational text. 5 - Copy only the files you intend to modify into a0/agents/your-agent/prompts/, and edit them there. Important: do not copy the entire a0/prompts folder into a0/agents/your-agent/prompts/. You don’t need all prompts whatever your use case is, because any file not present in your agent folder will fall back to the framework default. 6 - Once the new folder is created, your agent automatically appears in the Settings UI (first tab). The folder name becomes the agent name.

Agent Zero runs on Docker, that's the engine. But you shouldn't need to be a mechanic just to drive the car. Sneak peek at the upcoming A0 Launcher. One unified app to manage your instance, updates, and config. Docker stays under the hood, where it belongs.

@Erich Greenebaum and anyone else please let's coordinate a time to connect!~ Great discussion today thank you!

Sorry if this has been covered maybe needs a reboot: Is there some kind of disclaimer or indemnification needed to shield from lawsuits the use of A0 in malicious attacks? I don't think is necessary is it? I'm curious. I mean there is hacker stuff built into the system...soo...should be fine right?

Current experience: >> Install Agent Zero with Docker. >> New Chat >> Agent Zero: Hello! 👋, I'm Agent Zero, your AI assistant. How can I help you today? >> User: Hi, [...] >> Agent Zero: Recall memories extension error: [...]. Error [...] Looks familiar? Yes, we could just call people stupid or silly and send them to ask GPT how to solve this or to watch videos on how to start with Agent Zero. Or, we could implement the simplest, small local LLM setup within the image that would be able to run a predefined routine for the onboarding experience. Does this idea sound ridiculous? 🤔