Overview
arbor is an open-source data skill for Claude Code and compatible agents, published by K-Dense-AI. Its author describes it as: “Autonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and an evaluator, using Hypothesis Tree Refinement (HTR) from the Arbor paper. Use this whenever…”. The project has 46k stars on GitHub and is available under the MIT license. Add it to your setup with `git clone https://github.com/K-Dense-AI/scientific-agent-skills ~/.claude/skills/arbor`.
What arbor does
This skill runs an **Autonomous Optimization (AO)** loop: starting from an existing artifact and a measurable objective, improve it through many rounds of experiment and evaluation — without step-by-step human supervision and without overfitting to the feedback signal. It's the right tool when the bottleneck isn't writing one good change, but *organizing dozens of trials* so that lessons accumulate instead of evaporating.
Best for
What’s inside
The SKILL.md for arbor is organised into these sections:
- Overview
- When to use this skill
- The AO setup — pin this down first
- The coordinator loop
- 1. Observe
- 2. Ideate
- 3. Select
- 4. Dispatch
- 5. Backpropagate
- 6. Decide
- Finishing the run
- Principles that make this work (not rote rules)
Installation
Add arbor to your agent with:
git clone https://github.com/K-Dense-AI/scientific-agent-skills ~/.claude/skills/arborReview a skill’s source before installing it — skills can instruct your agent to run commands. This command comes from the skill’s public repository, which is the source of truth for setup steps. New to skills? Read how to install Claude Code skills.
Supported agents
Neither the skill nor its repository names a specific agent. It uses the SKILL.md format, which several agents can load — check the repository for any agent-specific requirements.
License
arbor is published under the MIT license.
Strengths
- Clear MIT license — safe to read and adapt within its terms
- Ships in K-Dense-AI/scientific-agent-skills, an established project with 45,887 GitHub stars
- Actively maintained — repository updated in the last 30 days
- Its repository gained 12,066 GitHub stars over the last 34 days
Author
Topics
Frequently asked questions
- What does arbor do?
- Autonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and an evaluator, using Hypothesis Tree Refinement (HTR) from the Arbor paper. Use this whenever someone wants to iteratively optimize something over many experiments without overfitting — e.g. "get my model's eval score up", "improve this agent/harness", "tune this pipeline", "beat the baseline on this benchmark", "run a search over approaches and keep the best", "do an MLE-bench / Kaggle-style optimization", or any long-horizon "make this artifact better and don't just memorize the dev set" task. Trigger it even when the user doesn't say "Arbor" or "hypothesis tree" but describes repeated experiment-and-evaluate loops, branching exploration of competing ideas, or worries about a dev/test gap. Runs Claude itself as the coordinator with subagent executors in isolated git worktrees; for the standalone `arbor` CLI tool see references/arbor-upstream.md.
- How do I install arbor?
- Run git clone https://github.com/K-Dense-AI/scientific-agent-skills ~/.claude/skills/arbor in your agent, then reload your skills. Review the source at https://github.com/K-Dense-AI/scientific-agent-skills before installing.
- Is arbor free to use?
- Yes. arbor is free and open source under the MIT license, so you can read, run, and adapt it within that license's terms.
- Where does arbor come from?
- arbor ships inside K-Dense-AI/scientific-agent-skills, a repository that contains 42 catalogued skills in total. The repository's 45,887 GitHub stars apply to that whole collection, not to this skill on its own.