research-judging

Evaluate AI-generated research outputs using structured JSON judgments.

13|Updated Jan 27, 2026

Search Tags:#research #schema #rubric #evaluate #auditable #structured-json

Ready to run

npx skills add https://github.com/Butanium/claude-lab --skill research-judging

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: research-judging
Source: https://github.com/Butanium/claude-lab/tree/main/plugins/clab/skills/research-judging
Command: npx skills add https://github.com/Butanium/claude-lab --skill research-judging

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Evaluate AI-generated research outputs using structured JSON judgments.

Core Features & Use Cases

Provides a repeatable judging pipeline with a clear rubric (criteria.md) and a strict JSON schema (schema.json).
Supports model-agnostic evaluation by applying consistent scoring across samples and exportable results.
Use Case: A team reviews hundreds of AI-generated drafts and needs auditable, comparable assessments.

Quick Start

Run claude -p on a directory of samples using the provided criteria and schema to generate structured judgments.

Frequently Asked Questions about research-judging

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

FAQPage Schema

How do I evaluate AI-generated research outputs with a structured rubric?▼

To evaluate AI-generated research outputs with a structured rubric, this Skill applies a defined criteria.md and a strict JSON schema to standardize scoring and enable auditable comparisons across multiple samples.

How does structured JSON judging work for multiple AI research samples?▼

Structured JSON judging works by enforcing a defined rubric and a strict JSON schema across multiple AI research samples, ensuring consistent scoring and enabling machine-validated, auditable comparisons between outputs.

Can I use a custom JSON schema to standardize scoring for AI drafts?▼

Yes, you can enforce a custom JSON schema defined in schema.json alongside a rubric in criteria.md to standardize scoring and ensure consistent, machine-validated results across hundreds of AI-generated drafts.

What is the best way to generate auditable comparisons across hundreds of research drafts?▼

The best way to generate auditable comparisons across hundreds of research drafts is to run a model-agnostic evaluation pipeline that applies a consistent scoring rubric and outputs structured JSON judgments.

Does this research judging pipeline work without external dependencies?▼

Yes, this research judging pipeline operates without external dependencies, relying solely on its internal criteria.md rubric and schema.json to produce structured JSON judgments for AI-generated research outputs.

How do I run automated research evaluation on a directory of samples?▼

You run automated research evaluation by executing the pipeline on a directory of samples using the provided criteria and schema, which generates structured JSON judgments with consistent scoring.