A new implementation of Microsoft SkillOpt shows prompt optimization as an instrumented workflow rather than a one-off rewrite.

The setup runs optimization loops with rollout, reflection, aggregation, selection, and validation-based gating, then compares the evolved skill against a baseline. That makes prompt improvement easier to measure and debug.

For teams building agent systems, the useful idea is not just better prompts. It is treating prompt and skill evolution as a controlled engineering process with metrics and failure analysis.