evaluator-optimizer

Pair a generator and evaluator to iteratively refine outputs.

2|Updated Jan 16, 2026

Search Tags:#automation #workflow #multi-agent #quality-assurance #generator #evaluator #iterative-improvement

Ready to run

npx skills add https://github.com/MysMon/Spec-Workflow-Toolkit --skill evaluator-optimizer

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: evaluator-optimizer
Source: https://github.com/MysMon/Spec-Workflow-Toolkit/tree/main/skills/workflows/evaluator-optimizer
Command: npx skills add https://github.com/MysMon/Spec-Workflow-Toolkit --skill evaluator-optimizer

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Iterative improvement pattern using evaluator-optimizer feedback loops between a generator and an evaluator to reach high-quality outputs.

Core Features & Use Cases

Generator-Evaluator dual agents for iterative refinement and quality assurance
Supports multi-round evaluation, scoring, and guardrails (max iterations, pass thresholds)
Integrates with Plan/Review/Implement workflows for documentation, code, and design tasks
Suitable for complex outputs like docs, designs, or algorithms

Quick Start

Run an evaluator-optimizer loop on an initial draft to iteratively refine quality until criteria are met.

Frequently Asked Questions about evaluator-optimizer

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

FAQPage Schema

What is an evaluator-optimizer feedback loop for multi-agent workflows?▼

An evaluator-optimizer feedback loop is a multi-agent architecture pairing a generator and an evaluator to progressively enhance output quality through iterative improvement rounds.

How do I use iterative refinement to improve code reviews and API design?▼

Iterative refinement improves code reviews and API design by running an initial draft through dual agents that score outputs against explicit thresholds over multiple rounds until quality criteria are met.

Can I set scoring thresholds and max iteration limits for automated quality assurance?▼

Yes, automated quality assurance supports explicit max-iteration limits and scoring thresholds to govern the generator-evaluator loop, returning the best-attempt output if thresholds are not reached.

What's the best way to automate documentation refinement with a dual-agent architecture?▼

Automating documentation refinement is best handled by a dual-agent architecture where independent tools handle generation and evaluation, applying multi-round scoring to iteratively improve complex drafts.

Does the generator-evaluator pattern work without external dependencies?▼

The generator-evaluator pattern works without external dependencies, using independent tools within the dual-agent architecture to execute the iterative improvement workflow and return refined outputs.

When should I avoid using iterative improvement for algorithm optimization?▼

Iterative improvement for algorithm optimization should be avoided when strict max-iteration limits cannot accommodate the required multi-round refinement cycles, returning only a best-attempt result instead.