harness-eval

Evaluate Claude Code harnesses across CLAUDE.md, hooks, settings, and skills.

Updated May 15, 2026

Search Tags:#python #claude code #codebase analysis #quality assessment #harness evaluation

Ready to run

npx skills add https://github.com/heyu-ai/yibi-stack --skill harness-eval-heyu-ai

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: harness-eval
Source: https://github.com/heyu-ai/yibi-stack/tree/main/tasks/harness_eval
Command: npx skills add https://github.com/heyu-ai/yibi-stack --skill harness-eval-heyu-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires python, click, pydantic, pathlib, json, os, datetime, math, re, subprocess, collections, json, yaml, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill assesses the readiness and quality of Claude Code harnesses, providing insights into their performance and potential issues.

Core Features & Use Cases

Harness Evaluation: Automatically scan Claude Code harnesses for various dimensions such as CLAUDE.md quality, hooks, settings, skills, and more.
Dimension Scores: Receive detailed scores for each evaluated dimension, highlighting strengths and areas for improvement.
Semantic Scoring: Leverage additional semantic scoring for more nuanced insights.

Quick Start

Run the 'harness-eval scan' command to evaluate the harness in the current directory.

Frequently Asked Questions about harness-eval

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

FAQPage Schema

How do I evaluate Claude Code harness readiness and quality?▼

To evaluate Claude Code harness readiness, run the 'harness-eval scan' command in your project directory to automatically assess CLAUDE.md quality, hooks, settings, and skills. This generates detailed dimension scores highlighting strengths and areas needing improvement.

What dimensions does a Claude Code harness quality assessment cover?▼

A Claude Code harness quality assessment covers multiple dimensions including CLAUDE.md file quality, hooks configuration, settings validation, and skills deployment. Each dimension receives a detailed score to highlight strengths and areas for improvement.

Do I need Python to run a Claude Code harness evaluation?▼

Yes, you need Python installed along with specific libraries such as click, pydantic, yaml, and pathlib to run harness evaluations. These dependencies are required for the tool to scan directories and perform semantic scoring.

How do I check if my CLAUDE.md configuration is ready for production?▼

To check if your CLAUDE.md configuration is ready, execute a harness scan to receive a semantic readiness score. The evaluation automatically checks your configuration quality and provides nuanced insights into potential performance issues.

What is the best way to score Claude Code hooks and settings?▼

The best way to score Claude Code hooks and settings is using an automated harness evaluation tool that scans your configuration files and assigns dimension scores. This approach leverages semantic scoring for nuanced insights into your setup.

Why does my Claude Code harness evaluation require subprocess and yaml libraries?▼

Your Claude Code harness evaluation requires subprocess and yaml libraries to scan configuration files, execute system checks, and parse YAML settings. These dependencies enable the automated assessment of hooks and settings across your project.