skills.rest
junhyekh
junhyekhCOMMUNITY
Explore all skills by this author →
GitHub

rlxp-audit-results

Audit reinforcement-learning improvement claims for validity across RLXP experiments.

1|Updated May 14, 2026
Search Tags:#reinforcement learning#guardrails#experiment audit#evaluation protocol#candidate validation#metric provenance
One-click install
npx skills add https://github.com/junhyekh/rlxp --skill rlxp-audit-results
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: rlxp-audit-results
Source: https://github.com/junhyekh/rlxp/tree/main/plugins/rl-experiment-assistant/skills/rlxp-audit-results
Command: npx skills add https://github.com/junhyekh/rlxp --skill rlxp-audit-results

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you decide whether an RLXP improvement claim is trustworthy before you promote a candidate or accept an autoloop result.

Core Features & Use Cases

  • Checks task and study identity so results are matched to the correct contract and scope.
  • Verifies metric provenance, held-out evaluation, checkpoint selection, guardrails, and output consistency.
  • Use it when reviewing a new experiment, validating an autoloop proposal, or rejecting a claim that relies only on training reward.

Quick Start

Use the rlxp-audit-results skill to audit the attached candidate, contract, metrics, run summary, and report claim for promotion readiness.

Frequently Asked Questions about rlxp-audit-results

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I validate reinforcement learning improvement claims before promoting a candidate model?▼

Reinforcement learning audit checks verify that evaluation protocols remain unchanged, held-out splits are properly maintained, checkpoint selection is valid, guardrails are enforced, and training-reward substitution is absent to confirm claim validity.

How do I audit autoloop outputs for task and study identity consistency?▼

Rejecting reinforcement learning claims based solely on training reward is necessary because training-reward substitution does not reflect true held-out evaluation performance, making the promotion claim invalid without proper metric provenance verification.

What is the best way to verify metric provenance in a reinforcement learning experiment audit?▼

To audit a reinforcement learning experiment for promotion readiness, attach the candidate, contract, metrics, run summary, and report claim, then verify task identity, metric provenance, held-out splits, checkpoint selection, and guardrails.

Can I use an experiment audit to check if held-out splits and evaluation protocols were changed?▼

A reinforcement learning experiment audit requires the candidate model, contract, metrics, run summary, and report claim to verify task identity, metric provenance, checkpoint selection, guardrails, and the absence of training-reward substitution.

Related Skills

Explore More Software Engineering Skills→
junhyekhjunhyekh
1

rl-experiment-loop

Coordinate reinforcement-learning experiment loops with validation, monitoring, and launch gates.

Community
Advanced
anderssvanderssv
14

kotlin-sum-types

Model Kotlin validation outcomes as sealed Valid/Invalid types with create() parsers.

Community
Advanced
junhyekhjunhyekh
1

rlxp-update-report

Update RLXP study reports from structured experiment evidence.

Community
Intermediate
junhyekhjunhyekh
1

rl-reward-curriculum-design

Route legacy RL reward and curriculum requests to RLXP intervention skills.

Community
Intermediate