rlxp-reward-reflection

What problem does it solve?

This Skill helps you explain why a reward-related reinforcement learning run succeeded or failed by grounding the analysis in local task metrics and reward-component trends instead of training reward alone.

Core Features & Use Cases

• Evidence-backed diagnosis: Detect saturated, inactive, dominating, or conflicting reward terms from local artifacts.
• Scope-aware analysis: Confirm the run belongs to the approved task or study contract before drawing conclusions.
• Next-step guidance: Decide whether the next move should be scalar reward tuning, reward-code changes, or a different validation path.
• Use case: After a reward-engineering run, use this Skill to compare reward behavior against the task metric and guardrails, then write a structured reflection for the next iteration.

Quick Start

Ask the assistant to analyze the local run artifacts for the active RLXP contract and produce a reward reflection JSON with the diagnosis and next recommended action.