Roni Kobrosly Ph.D.'s Website

A python package for agentic loop engineering that employs rigorous data science methods to evaluate candidate solutions.

A statistically-sound agentic loop-engineering framework. You give it a task description, a pile of gold-standard input/output examples, and a set of checks; it runs agentic loops (a strategy agent directing concurrent executor agents) that iteratively build and refine a solution — and it evaluates that solution the way a careful data scientist would: on a strict dev / validation / test split, so the score you see at the end is one you can actually trust.

The solution it produces is a portable artifact you can take away and use without RigorLoop:

• an executable Python script (reads your input on stdin, writes the output),
• an agent skill (a SKILL.md-style document, e.g. for Claude Skills), or
• a guidance file (an AGENTS.md/CLAUDE.md-style document for coding agents).

The GitHub repository can be found here.