Codify pass/fail criteria and baseline performance for Claude Code sessions.
npx skills add https://github.com/danielkhigay/gaz --skill eval-harness-danielkhigay
Please help me install this Agent Skill. Skill: eval-harness Source: https://github.com/danielkhigay/gaz/tree/main/.claude/skills/archived/eval-harness Command: npx skills add https://github.com/danielkhigay/gaz --skill eval-harness-danielkhigay
Formal evaluation framework for Claude Code sessions implementing eval-driven development (EDD) principles helps teams define, run, and track structured assessments of AI outputs, enabling early detection of regressions and clearer success criteria.
Define eval criteria, implement tests, and run the eval suite to produce a results report.
High-intent search queries and answers about installing and using this skill.
Regression testing for AI coding sessions is set up by codifying pass/fail criteria and baselining performance. You define deterministic graders to run repeatable checks across different prompts and model versions.
Eval-driven development evaluates Claude Code outputs by defining, running, and tracking structured assessments. It captures pass@k and pass^k metrics to establish clear success criteria and detect regressions early.
AI task evaluation supports three distinct grader types: code-based, model-based, and human. This flexible validation approach allows teams to apply deterministic checks tailored to specific output requirements.
Quality assurance metrics are tracked by storing standardized eval artifacts in .claude/evals. This enables reproducible baseline testing and consistent regression checks across varying model versions and prompts.
This evaluation framework operates without external dependencies. It relies entirely on implementing structured eval criteria and deterministic graders to produce reproducible baseline testing results.
The best way to codify pass/fail criteria is implementing structured capability and regression evals. This formal approach applies deterministic graders to validate AI outputs and generate standardized eval artifacts.
JamkrisAutomate eval-driven testing workflows for Gemini CLI sessions with deterministic graders.
SeanChenRAutomate evaluation and regression tracking for AI-assisted coding sessions.
dongitranDefine and run capability and regression evals for Claude Code development.
petrkindlmannValidate AI features by testing prompts, tool calls, and nondeterministic outputs.
