Vein05

Rharness — AI skill for Claude Code

AI community

An opinionated research harness for ML and NLP work with coding agents.

How to install Rharness

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Vein05/rharness and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Rharness does

An opinionated research harness for ML and NLP work with coding agents. One command, a disciplined workspace.

Alternatives in AI

ARIS Agent Guide — For AI agents reading this repo 6.2k ★
J Space Cognition Suite V3.7 — J-Space Cognition Suite V3.7 - AI cognitive-enhancement Skills based on Anthropic's J-space global workspace r 3k ★
Better Harness — Review this project's AI coding workflow with Better Harness and generate a durable report 2.1k ★

README

rharness

Cover made with the rharness figures plugin.

rharness is an opinionated research harness for ML and NLP work done with coding agents. One command gives you a workspace where every paper project has a contract the agent must obey, a record of every session, and a lint that tells you when the two drift apart. It is markdown and small scripts, not an app: the agent reads the files, you edit them, git tracks them.

The opinions

These are not defaults you configure. They are the product. Also at [vein05.github.io/rharness](https://vein05.github.io/rharness/#opinions).

A kill criterion before any spend. Every project starts with CHARTER.md: the core question, the result that means the paper exists, and the result that means stop.
Spec before code. The headline table is designed in spec/scoring.md before a single number exists, with its control row and its ceiling row. Code without a spec is a probe and never produces a paper number.
Never rewrite history. Handoffs, changelogs, and experiment reports are dated and append-only. Corrections are added on top with a date.
**Every number trace