validate-output

Validate HFI outputs against a 38-check framework with severity reporting.

Updated Aug 27, 2026

Search Tags:#quality #validate #report #output #checks #HFI #38-check

Ready to run

npx skills add https://github.com/MinhOmega/marlin-skill --skill validate-output-minhomega

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: validate-output
Source: https://github.com/MinhOmega/marlin-skill/tree/main/.claude/skills/validate-output
Command: npx skills add https://github.com/MinhOmega/marlin-skill --skill validate-output-minhomega

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Run quality checks against HFI output text. Reports violations by severity.

Core Features & Use Cases

38-check evaluation across prompts, evaluations, and final outputs.
Provides actionable fix suggestions and severity-based reporting.
Use case: Ensure final outputs adhere to safety, quality, and prompt guidelines before submission.

Quick Start

Run validate-output on the provided HFI output to generate a structured validation report.

Frequently Asked Questions about validate-output

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

FAQPage Schema

How do I validate the quality of HFI outputs before submission?▼

You can validate HFI outputs by running an automated 38-check quality gate across prompts, evaluations, and final outputs. This process generates a structured report detailing violations by severity and providing actionable fix suggestions to ensure adherence to safety and prompt guidelines.

What is a 38-check quality validation framework for output text?▼

A 38-check quality validation framework is an automated system that evaluates HFI output text against a rigorous set of safety, quality, and prompt guideline criteria. It categorizes any detected violations by severity level and provides actionable suggestions to fix the reported issues.

How do I apply quality checks across multi-turn feedback loops?▼

To apply quality checks across multi-turn feedback loops, run automated validation against the prompts, evaluations, and final outputs generated during the process. The validation system reports per-severity issues at each stage and returns actionable fix suggestions to refine the loop.

Does this output validation provide severity-based reporting for issues?▼

Yes, output validation provides severity-based reporting by categorizing all detected issues according to their severity level. It evaluates text against a 38-check framework and pairs each reported violation with actionable fix suggestions to guide the correction process.

What's the best way to ensure generated outputs adhere to safety and prompt guidelines?▼

The best way to ensure adherence to safety and prompt guidelines is to pass the generated text through an automated 38-check quality validation gate. This evaluates evaluations, prompts, and final outputs, reporting per-severity issues with actionable fix suggestions before submission.

Can I get actionable fix suggestions from an automated output validation report?▼

Yes, you can get actionable fix suggestions from an automated output validation report. When the 38-check framework detects violations across prompts, evaluations, or final outputs, it categorizes them by severity and provides specific recommendations to fix the flagged issues.