Can I use this for non-code content like documentation?

Yes, the skill includes specific patterns for content evaluation, such as documentation quality, marketing copy, and brand voice consistency.

How does the judge agent discover my rubrics?

The judge agent automatically scans the .claude/evals/ directory in your project for .yaml or .yml files containing your defined criteria.

What are quality gates in this context?

Quality gates are score thresholds, such as 75% for a PR review or 85% for production, that determine if the evaluated content meets your standards for that stage.

What is the difference between content and behavior evaluation?

Content evaluation checks static files like code style or copy, while behavior evaluation analyzes what Claude actually did, such as validating API responses or commit message quality.

How do I integrate this with other Claude plugins?

The skill provides programmatic integration patterns allowing other plugins to invoke the judge agent with specific rubrics and content for structured, automated feedback.

AI Quality Evaluation Patterns

Name: AI Quality Evaluation Patterns
Author: therealchrisrock

bytherealchrisrock

セキュリティとテスト

Automates quality assurance by providing standardized evaluation patterns and rubrics for code, content, and system behavior.

概要

The eval-patterns skill extends the eval-framework by offering a comprehensive library of patterns for integrating automated evaluations into your development workflow. It enables developers to implement project-local rubrics for content validation, security audits, and behavioral testing, ensuring that Claude's outputs consistently meet specific quality gates before deployment or commitment. By utilizing a proactive judge agent, it transforms subjective reviews into objective, structured feedback based on domain-specific best practices.

主な機能

0 GitHub stars
Standardized patterns for content, behavior, and combined evaluations
Domain-specific guidance for API design, security, and brand voice
Proactive judge agent integration for post-implementation reviews
Project-local rubric management using YAML-based discovery
Structured feedback loops with actionable improvement suggestions

ユースケース

Automating code security audits after implementing new authentication modules
Verifying marketing copy against brand voice guidelines within specific routes
Setting up pre-commit quality gates to ensure documentation and test coverage meet thresholds

概要

主な機能

0 GitHub stars
Standardized patterns for content, behavior, and combined evaluations
Domain-specific guidance for API design, security, and brand voice
Proactive judge agent integration for post-implementation reviews
Project-local rubric management using YAML-based discovery
Structured feedback loops with actionable improvement suggestions

ユースケース

Automating code security audits after implementing new authentication modules
Verifying marketing copy against brand voice guidelines within specific routes
Setting up pre-commit quality gates to ensure documentation and test coverage meet thresholds