# Harness OS

> Harness OS is a visible evaluation and repair loop for AI artifact generation. It accepts a goal, URL, reference image, or attached text, then creates a bespoke harness with weighted criteria, evaluator metadata, a design system, and repair vectors.

Canonical URL: https://harnessos.run
Repository: https://github.com/bonginkan/harness-os
AI index: https://harnessos.run/ai-index.json
Full LLM context: https://harnessos.run/llms-full.txt

## What It Does

- Routes artifact kinds: Auto, Web screen, Poster, Flyer, Image, Spec, Proposal.
- Critiques the input for domain, audience, intent, transformation, and observations.
- Synthesizes a harness with success/failure conditions, weighted criteria, evaluator mix, design system, and repair vector.
- Generates a baseline artifact, scores it with evidence, repairs weak criteria, and exports run evidence.
- Supports interactive web artifacts, visual image artifacts, and structured document artifacts.

## Disambiguation

Harness OS is not a generic chatbot or one-shot UI generator. It is an evaluator-optimizer workflow for making generated artifact quality inspectable through explicit criteria, evidence, repair decisions, and final scores.