Idea for improved agent to agent work

[timbo1975]

I love that agent0 can trigger subordinate agents and I think it would be amazing if using something like openrouter you would be able to say which LLM the agent uses. Why? because I have found thanks to A0 that different LLM's are better at certain tasks. For example, the developer agent could use Claude Opus and the researcher Gemini. I know this could be done via api and skill but imagine an agent that fails a coding task offloads it to another LLM and see if it can fix or find a solution.

One other suggestion is that when A0 gets stuck in a loop it would be good if you could set a step limit so that in the event of a loop it stops and requests user input to continue - this would stop burning credits unnecessarily.

I hope these suggestions are useful to the dev team and it hopefully helps with the awesome project.