What if we can able to add a self evolving loop for Micheal, that understands the user over time and adapts to their tone and usage?

[ajayjai30]

I have worked with hermes for a quite time, over the period, it updated itself many times, and adapted to my tone and everything, my style etc

It is not by the model, hermes has an internal methodology to improve itself over time

Eg: If hermes works for a long time as a stock trader, it will become stock trader over time

Also i have some other ideas too, i dont know whether the plugins or slash commands that is specialized for coding agent's is added or not

for eg:
Agent-skills(production grade slash commands),
caveman, which specializes in reducing context and also works good for multiagents,
ponytail, a lazy senior developer who specializes in building applications

if it is there it is great, if it is not, adding them as slash commands on default will help the developers much more because these are easily best skills that our god agent can directly inject onto the respected agents at the need

for eg: assuming claude code is the one which is buidng application on micheal's watch, micheal can detect that "ok claude code is building, lets inject ponytail-base inside the testing as skill file) so that it greatly helps the coding agent to build with what it has

And with the help of agent-skills, micheal can even become a good planning agent because of different planning methods inside the agent-skills(test-driven development, spec-driven development, idea-driven developemnt etc) which helps the micheal to manage and split the work (also caveman multiagent skill helps to use less token while doing it) reducing token cost and optimizing things at the same time.

If it is available it is great, but if possible we can take the self evolution concept idea from hermes agent to make munder difflin a better orchestrator over time

Also i have another thing that is needed, when added it makes much easier for developers as well,

Graphify, it can convert almost every code base into a queryable graph, that can be queried by the agent directly

I know almost every developer knows what is this, what it can do, so adding this inside munder diffliin with a direct node access to micheal will help micheal to also check for work if it has been done or not, or atleast if possible in case of multiple cli's are building multiple parts of the application, this single mcp can work with all of them at once so that the coding cli's can keep track and micheal can also check if everyone is doing their own work or not

helps more in parallel processing and token reduction.