Harness2 Claude Port — Development skill for Claude Code

harness2-claude-port
A Claude Code executor for Harness²

An independent port of Harness² - not from the original authors.

A **harness** is the system around a model that enables it to act as an agent: prompts, tools, memory, skills, orchestration, and more. **Harnessing** is the process of optimizing that system.

[Harness²](https://github.com/google-research/harness2) is Google Research's framework for **recursive agent harnessing** in an open world ([paper](https://arxiv.org/abs/2610.14224)). It connects two levels of recursion while keeping model weights fixed:

• Task-level recursion refines the execution harness for the current task through a contrastive Propose–Probe–Compose step: propose edits, probe their effects on the same task, and compose a refined harness from behavioral contrasts between executions, without ground-truth feedback.
• Harnessing-level recursion updates a persistent improvement harness with adopted harnesses and evidence-supported lessons from each task to guide future refinement.

This repository is an **independent port** that lets Harness² improve [Claude Code](https://github.com/anthropics/claude-code) agents, alongside upstream's OpenCode and Codex executors. It is not a Google project and is not endorsed by the Harness² authors. A sibling port does the same for Hermes Agent at [harness2-hermes-port](https://github.com/abhaymettu/harness2-hermes-port).

Contents

• [The idea in