Plan, execute, and evaluate AI evals with EZVals.
npx skills add https://github.com/camronh/EZVals --skill evals-camronh
Please help me install this Agent Skill. Skill: evals Source: https://github.com/camronh/EZVals/tree/main/ezvals/skills/evals Command: npx skills add https://github.com/camronh/EZVals --skill evals-camronh
Enables teams to plan, execute, and analyze evaluations of AI agents and LLMs using EZVals, turning evaluation planning into repeatable, auditable processes.
Initiate an evaluation planning session by outlining a target, dataset, and grader and then wire them into EZVals for execution.
High-intent search queries and answers about installing and using this skill.
To plan and run AI agent evaluations, define your target, dataset, and grader, then wire them into an evaluation framework like EZVals for execution and result comparison across runs.
An evaluation framework structures AI agent testing by pairing datasets of test cases with graders that score outputs. You need this process to turn LLM evaluation into repeatable, auditable workflows.
Comparing results across multiple AI evaluation runs involves executing experiments with defined targets and datasets, then analyzing the scored outputs side-by-side to review performance differences and patterns.
The best way to structure an LLM evaluation blueprint is to scaffold your evals, datasets, graders, and targets using structured metadata like YAML frontmatter, ensuring the process is repeatable and auditable.
No specific dependencies are required to execute AI evals. The evaluation framework operates independently, allowing you to plan, wrap targets, select datasets, and apply graders without external package constraints.
Yes, you can use custom scripts and references within your AI evaluation workflow. The framework supports optional scripts, references, and assets alongside the required structured metadata in the skill definition.
ResearAIStandardize a reference object for quest evaluation with plan and checklist requirements.
xqinagAnalyze ML experiment results to compute statistics and generate comparison tables.
ken-cavanagh-gleanEvaluate LLM outputs with direct scoring, pairwise comparison, and rubric generation.
jezwebConfigure structured JSON outputs and safety controls for Claude Agent SDK agents.
