skill-iter-tune

Workflow skill optimization uses an execute-evaluate-improve loop that repeatedly runs target skills, scores generated artifacts on clarity and correctness, and applies targeted fixes until quality thresholds are met or convergence occurs.

Evaluating chained skills requires structured phase loading and multi-dimensional scoring across execution rounds, applying rollback-safe modifications to skill definitions while collecting artifacts for iterative quality assessment.

Iterative skill tuning termination logic triggers based on defined quality thresholds, result convergence across evaluation rounds, or reaching maximum error limits during artifact generation.

Automated feedback loops suit workflow skills that function but require systematic tuning for better output reliability, safer behavior, and stronger performance across edge cases.

Yes, rollback-safe modification applies targeted improvements by selecting only the highest-impact changes suggested by the evaluator during artifact analysis, preventing regression in skill definitions.

Quality evaluation scores generated artifacts across five dimensions: clarity, completeness, correctness, effectiveness, and efficiency.