what is serenissima-harness fr?

This repository is a research workspace for building what its author calls a Universal Agent System, a setup where several AI agents work together on tasks instead of one AI model doing everything alone. The design is inspired by the historical Republic of Venice, where decisions were never made by a single ruler but through a group of councils working together. The workspace holds several kinds of material. There are written specification files that describe how the system should behave for different AI models, including Claude and Gemini. There is a folder of research papers, downloaded as PDFs, that the project draws ideas from, covering topics like coordinating many AI agents at once and building systems that can observe and improve themselves over time. There are also copies of other people's reference code that the project studies and builds on. The project describes a runtime layer that keeps track of ongoing work through a shared status file, task details, locks that prevent two agents from working on the same thing at once, and logs for reviewing what happened. It also includes integration folders for using this system directly inside Claude Code, Gemini, and Codex, each with their own agent roles such as a planner, a worker, and a verifier that checks the work. A coordinating AI agent is meant to read the main documentation first, then check the current status of the shared task board, and hand out available work to the other agents rather than doing every step in order by itself. Changes to the core specification files are meant to go through a review step rather than being made directly. This is an early-stage, experimental research project with very few stars, aimed at people already comfortable running multi-agent AI setups rather than newcomers. No license is stated in the README.