Convert successful agent execution traces into standardized and reusable skill modules.
Manually inspect logs and copy prompt/code snippets to write custom skill functions.
Manually reviewing logs and distilling prompts or constraints is tedious and error-prone.