gan-style-harness

Plan, generate, and evaluate applications using a multi-agent harness.

Updated Nov 19, 2025

Search Tags:#quality assurance #multi-agent #application development #GAN #AI harness

Ready to run

npx skills add https://github.com/Sake-Team/SmartSake --skill gan-style-harness-sake-team

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: gan-style-harness
Source: https://github.com/Sake-Team/SmartSake/tree/main/backup/skills/gan-style-harness
Command: npx skills add https://github.com/Sake-Team/SmartSake --skill gan-style-harness-sake-team

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires opus, playwright, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill solves the problem of creating high-quality applications with a multi-agent harness that separates generation from evaluation, using an adversarial feedback loop.

Core Features & Use Cases

Multi-Agent Harness: Separates generation from evaluation for quality assurance.
Adversarial Feedback Loop: Drives quality beyond what a single agent can achieve.
Use Case: Ideal for building applications from one-line prompts, frontend design, full-stack projects, and tasks requiring high visual quality.

Quick Start

Run the 'project:gan-build' command with your prompt, such as 'Build a project management app with Kanban boards, team collaboration, and dark mode'.

Frequently Asked Questions about gan-style-harness

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

FAQPage Schema

How does a multi-agent harness improve application development quality?▼

A multi-agent harness improves application development quality by separating generation from evaluation. The planner expands prompts into specifications, the generator implements features, and the evaluator tests the live application through an adversarial feedback loop to drive higher standards.

How do I build an application from a one-line prompt using an adversarial feedback loop?▼

Build an application from a one-line prompt by running the project:gan-build command. The harness planner expands the prompt into specifications, the generator implements the features, and the evaluator tests the live application to iterate on quality.

Do I need Playwright and Opus models to use this AI harness for full-stack projects?▼

Yes, this AI harness requires both Opus 4.6 models and Playwright. Opus models power the multi-agent harness components, while Playwright is specifically required for the evaluator agent to test the live application during development.

What is the best way to automate quality assurance for frontend design tasks?▼

Automate quality assurance for frontend design tasks by using a GAN-inspired multi-agent harness. The adversarial feedback loop between the generator and evaluator agents iteratively tests the live application, achieving higher visual quality than a single agent.

Can I use this multi-agent harness for tasks beyond full-stack application development?▼

Yes, beyond full-stack application development, the multi-agent harness is ideal for building applications from one-line prompts, frontend design, and tasks requiring high visual quality. The planner, generator, and evaluator structure adapts to various development scopes.