eval-harness
A formal evaluation framework for Claude Code implementing eval-driven development and reliability metrics.
Install
npx skills add https://github.com/affaan-m/everything-claude-code --skill eval-harnessStats
| Total installs | 1,576 |
| Weekly installs | 1.6K |
| GitHub stars | 112.5K |
| First seen | Jan 26, 2026 |
| Source | @affaan-m/everything-claude-code |
Summary
- Defines capability and regression evals with pass/fail criteria before implementation, treating evals as unit tests for AI-assisted workflows
- Supports three grader types: code-based (deterministic checks via bash/grep), model-based (Claude-as-judge), and human review for manual adjudication
- Tracks reliability with pass@k metrics (success within k attempts) and pass^k (all k trials succeed), with recommended thresholds of pass@3 ≥ 90% for capabilities and pass^3 = 100% for regressions
- Integrates into Claude Code workflow with commands to define evals before coding, check status during implementation, and generate reports post-completion
- Stores eval definitions, run history, and baselines in .claude/evals/ directory for version control alongside code
Tags
Related skills
| Skill | Installs | vs eval-harness |
|---|---|---|
| frontend-testing-best-practices | 1,457 | -119 |
| agent-device | 1,710 | +134 |
| springboot-tdd | 1,748 | +172 |
| n8n-validation-expert | 1,655 | +79 |
| swift-testing-expert | 1,781 | +205 |
FAQ
- How many installs does eval-harness have?
- eval-harness has 1,576 total installs and 1.6K installs this week.
- Where is eval-harness hosted?
- eval-harness is published by @affaan-m/everything-claude-code at https://github.com/affaan-m/everything-claude-code.
- How many GitHub stars does eval-harness have?
- eval-harness has 112.5K GitHub stars.
- When was eval-harness first indexed?
- OrangeBot.AI first indexed eval-harness on Jan 26, 2026.
- How do I install eval-harness?
- Run: npx skills add https://github.com/affaan-m/everything-claude-code --skill eval-harness