YC internal work with agents developed an alternative escape hatch approach: giving the LLM a 'debug_info' parameter in the response format that allows it to complain to the developer about confusing or underspecified information, and this debug_info becomes a to-do list for the agent developer to fix.

factualpending

Speaker

Jared

Evidence Quote

“We came up with a different way which is in the response format to give it the ability to have part of the response be essentially a complaint to you the developer”

Source

State-Of-The-Art Prompting For AI Agents— Y Combinator
Created: 8/11/2026, 7:07:30 AM

Your notes

Loading notes...