Introducing self-improvement loops.
The concept is simple: what if agents could improve Skills by reviewing past conversations?
Here's the three step loop:
- Score conversations from criteria you define
- Isolate failures
- Generate skill improvements