SafeDusk

mplewis thanks for helping to point those out!

I did! Thanks for responding and continue to do your great work, I'm a fan as a fellow Singaporean!

In addition to having long system prompts, you also need to provide agents with the right composable tools to make it work.

I’m having reasonable success with these seven tools: read, write, diff, browse, command, ask, think.

There is a minimal template here if anyone finds it useful: https://github.com/aperoc/toolkami

I’m having reasonable success specifically with Gemini model using only 7 tools: read, write, diff, browse, command, ask, think.

This minimal template might be helpful to you: https://github.com/aperoc/toolkami

This is why I tend to build a simple version of a complex tech just to feel what I am getting into - https://github.com/aperoc/toolkami (a minimal AI agent) as an example!

In a week if there is demand for it!