Harness Score Live Demo Teaches Measuring AI Coding Harnesses at Maturity Level L3
An open-source repository called Harness Score Live Demo offers an interactive, runnable lesson on evaluating how well a codebase guides and constrains AI coding assistants. The project is an MIT-licensed, Brazilian Portuguese presentation backed by real repository artifacts, and it requires only Git, Node.js 22, npm, and a terminal, with no API key needed. Its checks cover linting, type checking, tests, and builds, and a Harness Score scan of the current checkout reports 91 of 108 points, or 84%, at level L3. The lowest category is Hooks and Guardrails, which scored 0 of 14 because the demo deliberately omits agent-runtime hooks to illustrate the difference between written rules and enforced behavior. The author notes that scores reflect only this repository at the scanned revision and may change with different tool versions or scope.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.


Discussion (0)
Log in to join the discussion and vote.
Log in