CtrlK
BlogDocsLog inGet started
Tessl Logo

arbor

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.

77

Quality

96%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

92%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A high-quality, action-oriented body: concrete tree.py commands, an explicit six-step loop with a held-out merge-gate validation checkpoint, and clean one-level-deep references to real bundle files. The only weakness is minor rhetorical padding that slightly inflates the token budget.

Suggestions

Trim rhetorical asides like "This is what turns trial-and-error into cumulative, auditable research" and "Getting this right matters more than any later decision" that restate the value proposition rather than instruct.

The parenthetical ablation numbers in step 5 (54.5% vs 63.6% vs 81.8%) justify insight propagation but could be condensed to one sentence to save tokens while keeping the evidence.

DimensionReasoningScore

Conciseness

Mostly efficient and the methodology explanations (HTR, the ablation numbers) are justified because the topic is niche rather than common knowledge, but a few rhetorical asides — "This is what turns trial-and-error into cumulative, auditable research", "Getting this right matters more than any later decision" — could be trimmed. Not a 5 because not every token earns its place; not a 3 because the padding is minor rather than pervasive.

4 / 5

Actionability

Fully executable, copy-paste-ready `python scripts/tree.py` commands throughout (init, observe, cycle, add-node, set-status, set-evidence, propagate, prune, merge) with realistic arguments, plus a concrete four-field executor return contract — matching the copy-paste-ready anchor.

5 / 5

Workflow Clarity

A clearly sequenced six-step coordinator loop (Observe→Ideate→Select→Dispatch→Backpropagate→Decide) with an explicit validation checkpoint — the held-out merge gate run in a fresh worktree — and a feedback loop (gate rejection recorded as a lesson rather than silently promoting), so the batch/destructive cap does not apply.

5 / 5

Progressive Disclosure

SKILL.md is a well-signaled overview; deeper material lives one level deep in real reference files (htr-methodology.md, executor-brief.md, report-template.md, arbor-upstream.md, all confirmed present) cited inline at the relevant step and summarized in a final Reference files section.

5 / 5

Total

19

/

20

Passed

Description

100%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

An exemplary description: third-person voice, concrete actions, explicit what-and-when with abundant natural trigger phrases, and a distinct HTR/Arbor niche that routes away from adjacent skills. It tells Claude exactly when to fire without over-claiming.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — "Autonomously improve a real artifact...against an objective and an evaluator", "Runs Claude itself as the coordinator with subagent executors in isolated git worktrees" — across a comprehensive set of artifact types (code, recipe, harness, pipeline, prompt), matching the comprehensive-coverage anchor.

5 / 5

Completeness

Explicitly answers both what ("Autonomously improve a real artifact...using Hypothesis Tree Refinement") and when ("Use this whenever someone wants to iteratively optimize something over many experiments without overfitting — e.g. ...") with concrete trigger phrases, matching the 5-anchor.

5 / 5

Trigger Term Quality

Comprehensive natural-language triggers users would actually say — "get my model's eval score up", "tune this pipeline", "beat the baseline on this benchmark", "do an MLE-bench / Kaggle-style optimization" — plus synonyms for the same intent and guidance to trigger even when "Arbor" isn't named.

5 / 5

Distinctiveness Conflict Risk

Occupies a clear niche (HTR from the Arbor paper, dev/test held-out merge gate) with distinct triggers and explicit routing away from open-ended ideation skills, giving minimal conflict risk.

5 / 5

Total

20

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
K-Dense-AI/scientific-agent-skills
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.