evaluate

Compare reference and development products via scripted tests and screenshots.

8|1|Updated Mar 22, 2026

Search Tags:#qa #comparison #evidence #playwright #baseline #evaluate

Ready to run

npx skills add https://github.com/zxdxjtu/harness --skill evaluate-zxdxjtu

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: evaluate
Source: https://github.com/zxdxjtu/harness/tree/main/skills/evaluate
Command: npx skills add https://github.com/zxdxjtu/harness --skill evaluate-zxdxjtu

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Adversarial evaluation helps verify that the development product diverges from a reference baseline and documents concrete differences.

Core Features & Use Cases

Automated feature-by-feature comparison between reference and dev products.
Evidence generation with screenshots and logs to support scoring.
Structured evaluation workflow including baselines, specs, and a final evaluation report.

Quick Start

Compare the reference and development products step-by-step and generate a complete evaluation report.

Frequently Asked Questions about evaluate

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

FAQPage Schema

How do I compare a development product against a reference baseline?▼

To compare a development product against a reference baseline, you can perform automated feature-by-feature comparison, capture screenshots, and generate a structured evaluation report with dimension scores and actionable fix tasks.

Can I use Playwright for automated feature comparison and evidence generation?▼

Yes, scripted interaction tests using Playwright can apply to features loaded from the baseline and specs, performing evidence capture and screenshots to support objective scoring of the development product.

What is objective dev-vs-baseline evaluation in software engineering?▼

Objective dev-vs-baseline evaluation is an adversarial process that identifies concrete differences between a reference baseline and a development product, documenting observations through scripted tests and generating a structured evaluation report.

How do I generate an evaluation report with dimension scores and fix tasks?▼

You generate an evaluation report by applying scripted interaction tests and evidence capture to baseline features, which produces dimension scores, feature-level findings, and actionable fix tasks in the final structured document.

Does baseline comparison require loading specific specs and features?▼

Yes, baseline comparison requires loading features from the baseline and specs to accurately identify differences, apply scripted interaction tests, and document objective observations for the final evaluation report.

What is the best way to document concrete differences between dev and baseline products?▼

The best way to document concrete differences is through automated feature-by-feature comparison with evidence generation, capturing screenshots and logs to support scoring within a structured evaluation workflow.