evaluator-optimizer

Automates generator-evaluator loops to refine outputs against a rubric and quality gates.

1|Updated May 21, 2026
One-click install
npx skills add https://github.com/hiddink-ai/hiddink-harness --skill evaluator-optimizer-hiddink-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: evaluator-optimizer
Source: https://github.com/hiddink-ai/hiddink-harness/tree/main/templates/skills/evaluator-optimizer
Command: npx skills add https://github.com/hiddink-ai/hiddink-harness --skill evaluator-optimizer-hiddink-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill solves the problem of inconsistent AI output quality by implementing a rigorous, multi-turn feedback loop that ensures generated content meets specific, measurable standards before reaching the user.

Core Features & Use Cases

  • Iterative Refinement: Automatically cycles between a generator and an evaluator agent to improve output quality based on a defined rubric.
  • Configurable Quality Gates: Supports multiple validation strategies including score thresholds and majority-pass logic to ensure reliability.
  • Use Case: Use this for high-stakes tasks like architectural design, security-critical code generation, or complex documentation where a single-pass generation is insufficient.

Quick Start

Invoke the evaluator-optimizer skill to refine the current architectural plan using the standard quality rubric.

Frequently Asked Questions about evaluator-optimizer

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

FAQPage Schema
How do I automate quality assurance for critical software architecture tasks?

Automating quality assurance for software architecture requires an iterative generator-evaluator loop that refines output against a structured rubric until it passes defined quality gates. This approach ensures high reliability for critical tasks.

What is an iterative generator-evaluator loop for output refinement?

An iterative generator-evaluator loop is a multi-turn feedback mechanism where a generator creates content and an evaluator scores it against a rubric. The cycle repeats automatically until output meets score thresholds or majority-pass logic.

How do I configure quality gates to minimize token waste during AI refinement?

Configuring quality gates to minimize token waste involves setting specific validation strategies like score thresholds and majority-pass logic within an agent orchestration workflow. This structure manages refinement cycles efficiently.

When do I need a multi-turn feedback loop for security auditing or technical documentation?

You need a multi-turn feedback loop for security auditing or technical documentation when single-pass generation is insufficient and tasks demand rigorous validation. High-stakes domains require iterative refinement to meet measurable standards.

Can I use score thresholds and majority-pass logic for complex code generation validation?

Yes, you can apply score thresholds and majority-pass logic as configurable quality gates to validate complex code generation. These strategies ensure the generated code passes rigorous validation before delivery.

What are the limitations of using automated rubric-based refinement for workflow automation?

A limitation of automated rubric-based refinement is that it requires a predefined structured rubric and quality gate configuration to function. Without these inputs, the agent orchestration cannot effectively evaluate or refine output.