Skill Detail

Run TypeScript-native LLM eval suites with Evalite

Use Evalite when a TypeScript team needs repeatable Vitest-style eval runs for LLM apps before prompt, model, or workflow changes ship.

Code Quality & ReviewMulti-Framework
Code Quality & Review Multi-Framework Published
โญ 1.6k GitHub stars โฌ‡ 1.1M/wk npm
INSTALL WITH ANY AGENT
npx skills add agentskillexchange/skills --skill run-typescript-native-llm-eval-suites-with-evalite Copy
Works best when you want a reusable capability, not another fragile one-off prompt.
At a glance
Tools required
Node.js, a TypeScript project, Evalite npm package, model/provider credentials as required by the evaluated app
Install & setup
Install Evalite in a TypeScript project with `npm install evalite` or the package manager used by the repo, then add eval files, datasets, and scorers before running the Evalite test command documented upstream.
Author
Matt Pocock
Publisher
Individual
Last updated
Jun 18, 2026
Quick brief

Evalite is a TypeScript-native evaluation runner for LLM-powered apps. It is built around test-like eval files, datasets, scorers, and local reports so teams can treat model behavior changes like repeatable software tests.

How it works

What this skill actually does

Use this skill when an operator needs to define example cases, run an LLM or agent function against them, score outputs, inspect failures, and compare changes before altering prompts, models, retrieval behavior, or agent logic. It is especially useful when the surrounding application and CI workflow already live in TypeScript.

The boundary is repeatable eval execution for TypeScript LLM applications. It is not a general LLM observability platform, not a hosted product card, and not a generic testing library listing; the operator workflow is to install Evalite, write eval cases and scorers, run the eval suite, and review the report as a release gate.

Source evidence: upstream describes Evalite as a TypeScript-native, Vitest-based eval runner for LLM-powered apps, with exact npm package adoption mapped to the same GitHub repository.