rlxp-analyze-run

What problem does it solve?

This Skill helps you evaluate a completed reinforcement-learning run with evidence instead of guesswork, so you can tell whether a baseline or candidate truly improved and why.

Core Features & Use Cases

• Compares primary, guardrail, and diagnostic metrics against the current incumbent.
• Classifies the run as accept, reject, inconclusive, or debug and records the failure mode.
• Produces reward-focused reflections when needed, including saturation, domination, and suspected hacking signals.
• Use it after a training or evaluation job finishes, when a study needs a decision, or when you want to update the run evidence trail.

Quick Start

Use the rlxp-analyze-run skill to review the completed run against the active contract and determine the correct decision outcome.