mt-improve-skill

To empirically improve AI prompts, you evaluate them through unbiased execution scenarios, compare outcomes against predefined metrics, and iteratively apply targeted revisions until reliability converges.

Fixing ambiguous agent instructions requires structured failure analysis to trace execution errors, followed by controlled prompt revisions guided by measurable evaluation criteria to eliminate unclear guidance.

Yes, empirical testing applies to slash commands, task prompts, and skill definitions by running scenario-based evaluations to measure performance and iteratively refine instructions for consistent results.

Iterative instruction refinement works by executing scenario-based tests, analyzing failure phases structurally, and applying targeted prompt improvements repeatedly until the agent's output satisfies predefined metrics.

AI prompts generate inconsistent results when instructions contain ambiguity, requiring scenario-based evaluation and structured failure analysis to identify unclear guidance and apply controlled revisions for reliable outcomes.

Replace subjective prompt reviews with empirical evaluation when generated results become inconsistent and the root cause is unclear, requiring repeatable execution scenarios and metric tracking to achieve robust improvement.