All tiny agents
Permissions

Harness Decision Boundary

The model proposes. The harness decides. The log remembers.

What it does

A small TypeScript agent harness that separates tool visibility, permission, loop bounds, and event history. Based on the Module 1 mental model in Ship a Safe Agent. Includes 16 tests.

The guardrail

Tool visibility, permission checks and loop bounds live in the harness, not the model.

harnesspermissionsloop boundsevent log

Run it

git clone https://github.com/bobbyhalljr/tiny-harness-decision-boundary.git && cd tiny-harness-decision-boundary
npm test
npm run demo
Repo
tiny-harness-decision-boundary
Language
TypeScript
Stars
0
Price
Free