All tiny agents
Evals

Agent Evals

Did it stay in scope, act without approval, or hide what it did?

What it does

A scope eval for recorded agent runs. It asks three questions of every run, scores each judge against human labels, and fails the build if recall drops. Example data.

The guardrail

A build that fails when scope-eval recall drops.

evalsscopeCI

Run it

git clone https://github.com/bobbyhalljr/tiny-agent-evals.git && cd tiny-agent-evals
npm install
npx tsx evals.ts
Repo
tiny-agent-evals
Language
TypeScript
Stars
0
Price
Free