gan-style-harness
CommunityBuild production-quality apps with a GAN-style loop
System Documentation
What problem does it solve?
The GAN-style harness addresses the common failure mode where a single agent both generates and evaluates work, which produces optimistic self-assessments and low-quality outputs. It separates planning, generation, and ruthless evaluation into distinct agents so iterative feedback drives the system toward production-quality applications rather than mediocre prototypes.
Core Features & Use Cases
- Multi-agent architecture: Clear separation of Planner, Generator, and Evaluator roles to expand specs, implement features, and test live apps.
- Automated live evaluation: Uses Playwright-driven interaction and a weighted rubric (design, originality, craft, functionality) to score and provide actionable feedback.
- Iterative improvement loop: Configurable max iterations, pass thresholds, sprint contracts, and environment-driven behavior for continuous generate-test cycles.
- Use Cases: End-to-end full-stack app builds, high-fidelity frontend design workflows, and projects requiring production-grade polish rather than quick fixes.
Quick Start
Run the gan-style-harness to plan, generate, and evaluate a full-stack project until the weighted rubric score meets your pass threshold.
Dependency Matrix
Required Modules
None requiredComponents
Standard package💻 Claude Code Installation
Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.
Please help me install this Skill: Name: gan-style-harness Download link: https://github.com/Maelwalser/claude-config/archive/main.zip#gan-style-harness Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
Agent Skills Search Helper
Install a tiny helper to your Agent, search and equip skill from 471,000+ vetted skills library on demand.