gateway-debug

Diagnose G0-G4 gateway failures using MQS zero-scores, evidence JSON, and stderr.

1|Updated Jan 29, 2026
One-click install
npx skills add https://github.com/paiml/apr-model-qa-playbook --skill gateway-debug
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gateway-debug
Source: https://github.com/paiml/apr-model-qa-playbook/tree/main/.claude/skills/gateway-debug
Command: npx skills add https://github.com/paiml/apr-model-qa-playbook --skill gateway-debug

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you understand and resolve critical failures (G0-G4) that occur during the model qualification process, which can prevent a model from achieving a good score.

Core Features & Use Cases

  • Root Cause Analysis: Interprets error messages, evidence JSON, and stderr to pinpoint the exact cause of gateway failures.
  • Failure Triage: Guides you through a decision tree to quickly identify the failing gateway (G0-G4).
  • Specific Fixes: Provides targeted solutions for common issues like config.json mismatches, layout violations (LAYOUT-002), garbage output, and crashes.
  • Use Case: If your model qualification run fails with a zero score, this Skill will help you analyze the evidence to determine if the failure is due to incorrect model integrity (G0), loading issues (G1), inference problems (G2), stability (G3), or output quality (G4), and guide you toward a fix.

Quick Start

Use the gateway-debug skill to diagnose a G0-INTEGRITY failure in your model qualification run.

Frequently Asked Questions about gateway-debug

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

FAQPage Schema
How do I fix a zero score on my model qualification run?

A zero score on your model qualification run indicates a critical G0-G4 gateway failure. Diagnose the exact cause by interpreting evidence JSON, stderr output, and MQS zero-scores to identify issues like loading errors, crashes, or output quality failures.

What causes LAYOUT-002 violations during model qualification?

LAYOUT-002 violations during model qualification occur when output formatting does not meet structural requirements. Pinpoint these specific layout failures by analyzing the evidence JSON generated during the G4 output quality gateway checks.

How do I diagnose a G0-INTEGRITY failure in model qualification?

Diagnose a G0-INTEGRITY failure by checking for config.json mismatches and incorrect model integrity. Analyzing stderr output and evidence JSON helps pinpoint contract invariant violations (I-1 through I-5) that block model loading.

What is the best way to troubleshoot garbage output in model qualification?

Troubleshoot garbage output in model qualification by identifying GarbageOracle patterns in the inference results. This analysis determines if the issue stems from G2 inference problems or G4 output quality failures, guiding targeted fixes.

How do I interpret MQS zero-scores when troubleshooting gateway failures?

Interpret MQS zero-scores by mapping them to specific G0-G4 gateway failures. A decision tree helps quickly identify whether the zero-score originates from model integrity, loading issues, inference problems, stability, or output quality.

Why does my model qualification crash with a G3 stability failure?

A G3 stability failure means your model qualification run crashed during execution. Perform crash analysis on the stderr output to determine if the instability is caused by memory issues, contract invariant violations, or conversion diagnostics errors.