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 exper
供给资产档案
研究与知识工作
Deep research, source comparison, literature review, RAG, knowledge search, and reports.
场景
研究 Agent
I need my agent to research a topic, compare sources, and produce a concise report.
适配 Agent
Claude Code + CLI + Codex
适用于 Codex、Claude Code、Cursor、CLI 或自定义 Agent。
安装
就绪
npx skills add K-Dense-AI/scientific-agent-skills --skill arbor
维护状态
新鲜
距上次推送 2 天
风险
可安全尝试
可用元数据中未发现重大风险信号
GitHub 质量
34K
92/100 质量 · 84/100 信任
覆盖标签
审查说明
可用元数据中未发现重大风险信号
Agent 采用评分卡
一眼查看信任、审计与安装准备度
这些分数综合公开仓库元数据、OpenAgentSkill 审查信号、维护新鲜度与安装准备度。它用于候选筛选,不替代人工审查。
质量
优秀高置信候选,具有较强的采用度与健康维护信号。
信任
审查后安装适合加入候选清单,但 Agent 在运行前应审查审计说明、安装策略和结果证据。
审计
可安全尝试对安装准备度、安全元数据、维护情况与采用风险的机器可读审查。
OpenAgentSkill 信任评分 v5
安装前需人工审查
在人工审查或沙盒验证后作为首选候选。
Stars
34K 个 GitHub Stars
仓库活跃度
34K 个 Star,3.3K 个 Fork
维护状态
距上次推送 2 天
许可证
MIT license
安装
npx skills add K-Dense-AI/scientific-agent-skills --skill arbor
安装安全性
标准软件包或运行时安装路径
权限范围
shell or command execution, filesystem or document access
Agent 结果
暂未有 Agent 结果数据
文档
Usable metadata, review docs
风险摘要
低元数据风险
- 可用元数据中未发现重大信任警告
安装准备度
安装路径可用
- 安装路径可用
- 仓库证据可用
- 已声明许可证
- 暂无 Agent 验证结果证据
Agent 可读元数据
这个 Skill 的机器可读决策数据。
使用此区块或内嵌 JSON 判断 Agent 是否应安装该 Skill、选择替代方案,或先请求人工审查。
适用任务
- 研究 Agent 工作流
- Claude Code 团队
- 重视 GitHub 采用信号的团队
- 检索来源
适用 Agent
安装决策
- 命令
- npx skills add K-Dense-AI/scientific-agent-skills --skill arbor
- 策略
- 审查
- 人工审查
- 是
信任与风险
- 信任
- 79/100
- 审计
- 89/100
- 风险级别
- 可安全尝试
结果闭环
- 端点
- /api/agent/outcome
- 事件 ID
- resolve
- 结果
- 5
不适用场景
- 需要厂商支持 SLA 的团队
- 没有内部安全审查的高合规环境
- 当前元数据中未发现重大风险信号
- 高风险权限提示:Shell 或命令执行
- 可用元数据中未发现重大信任警告
替代 Skill
Last30days Skill
53.5K Stars
npx skills add mvanhorn/last30days-skill -g
替代 Skill
Academic Research Skills
38.4K Stars
npx skills add Imbad0202/academic-research-skills
替代 Skill
GPT Researcher
28.0K Stars
npx skills add assafelovic/gpt-researcher
替代 Skill
DeepResearch
19.8K Stars
npx skills add Alibaba-NLP/DeepResearch
Agent 安全 v2
61/100 · 安装前审查
可用候选,但 Agent 在安装前应展示权限与审计说明。
在真实工作区安装前需要人工批准。
高
Shell 或命令执行
Skill 元数据引用了终端、CLI、Shell、子进程或命令执行工作流。
中
网络访问
Skill 可能访问远程页面、API、仓库或外部服务。
中
文件系统访问
Skill 可能读取或写入项目文件、文档、生成产物或本地工作区状态。
- 高风险权限提示:Shell 或命令执行
安装目标
在你的 Agent 工作流中安装此 Skill
通过公开安装端点获取命令、安全清单、目标提示词和该 Skill 的规范链接。
OpenAgentSkill CLI
Resolve policy, run the source installer safely, and report a verified install receipt.
$ npx --yes https://github.com/Leon-Drq/openagentskill/releases/download/cli-v0.2.1/openagentskill-0.2.1.tgz install k-dense-ai-arborAgent 解析计划
让 Agent 在安装前验证匹配度。
Resolve API 返回首选 Skill、替代方案、安全策略、审计说明、安装目标和可直接执行的提示词,无需抓取此页面。
打开 JSON
/api/agent/resolve?task=Use%20arbor%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Resolve 文本
/api/agent/resolve?task=Use%20arbor%20for%20an%20agent%20workflow&agent=codex&max_risk=medium&format=text
安装交接
/api/skills/k-dense-ai-arbor/install
Agent 应检查
- 从 Resolve API 检查任务匹配与替代方案。
- 检查审计评分、信任评分和安全策略警告。
- 检查 Codex、Claude Code、Cursor 或 CLI 的安装目标兼容性。
复制提示词
Task: Use arbor in this workspace.
Resolve first: https://www.openagentskill.com/api/agent/resolve?task=Use%20arbor%20for%20an%20agent%20workflow&agent=codex&max_risk=medium
Review install handoff: https://www.openagentskill.com/api/skills/k-dense-ai-arbor/install
Install command: npx skills add K-Dense-AI/scientific-agent-skills --skill arbor
Before running it, summarize audit warnings, required permissions, and the fallback skill if install is risky.Agent 交接
把安装路径交给 Agent,而不是再给一个目录页。
通过公开安装端点获取命令、安全清单、目标提示词和该 Skill 的规范链接。
安装交接
/api/skills/k-dense-ai-arbor/install
LLM 文本格式
/api/skills/k-dense-ai-arbor/install?format=text
寻找替代方案
/api/skills/search?q=arbor&limit=3
Agent 提示词
Use arbor for this task. Review https://www.openagentskill.com/api/skills/k-dense-ai-arbor/install, then install with: npx skills add K-Dense-AI/scientific-agent-skills --skill arborRegistry 元数据
用于自动选择 Skill 的 Agent 可读档案。
本页通过 Registry API 提供相同的决策、信任、审计、场景和安装信号,让 Agent 无需抓取界面即可排序。
Agent 决策面板
适合 研究 Agent 的首选
将其作为优先候选,再在你的 Agent 环境中验证 README 与安装路径。
栈中角色
首选
主要匹配
研究 Agent
信任标签
可用于生产
安装路径
命令已就绪
适用场景
- 研究 Agent 工作流
- Claude Code 团队
- 重视 GitHub 采用信号的团队
证据
- 33,974 个 GitHub Stars
- 仓库近期活跃
- 已提供安装命令或 GitHub 仓库
- 92/100 质量档案
- 19 个 OpenAgentSkill 交互事件
先审查
- 当前元数据中未发现重大风险信号
实施路径
- 1在沙盒 Agent 中安装它,并端到端完成一次研究 Agent任务。
- 2Compare output quality, latency, and failure behavior against at least one alternative.
- 3Promote it into production only after reviewing repository permissions, license, and maintenance signals.
信任档案
审查后安装
适合加入候选清单,但 Agent 在运行前应审查审计说明、安装策略和结果证据。
GitHub 采用度
通过34K 个 GitHub Stars
Star/Fork 活跃度
通过34K 个 Star,3.3K 个 Fork; 当前元数据中没有议题活跃度信息
近期维护
通过距上次推送 2 天
许可证清晰度
通过MIT license
积极信号
- AI 审查已通过
- 安装路径可用
- 仓库证据可用
- 近期维护的仓库
- Large GitHub adoption signal
- 安装命令未发现明显高风险模式
- 结果闭环已就绪,但需要首次真实 Agent 运行
安装前审查
- 暂未有真实 Agent 结果报告
- 无人值守安装前需要人工审查
建议操作
在人工审查或沙盒验证后作为首选候选。
质量档案
优秀 适用于 Agent 工作流的候选
高置信候选,具有较强的采用度与健康维护信号。
工作流匹配
在这些场景使用此 Skill
Investigate faster
Research agents
I need my agent to research a topic, compare sources, and produce a concise report.
Search private knowledge
RAG and knowledge
I need my agent to build a RAG workflow over documents and retrieve reliable context.
Operate web apps
Browser automation
I need my agent to control a browser, fill forms, and verify web app workflows.
工作流匹配
加入完整工作流
Find, compare, and synthesize
Research report agent
A workflow for agents that gather sources, compare claims, summarize long material, and draft useful research briefs.
Ingest, retrieve, and cite
RAG knowledge base
A workflow for document-heavy agents that ingest files, create searchable knowledge, retrieve relevant context, and answer with grounded sources.
Operate and verify web apps
Browser QA agent
A workflow for agents that navigate products, fill forms, take screenshots, and verify real user flows across web applications.
替代方案短名单
安装前对比
可能适合该任务的相近 Skill。
Last30days Skill
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
Academic Research Skills
Academic Research Skills for Claude Code: research → write → review → revise → finalize
GPT Researcher
Run autonomous deep research over web and local sources
DeepResearch
Tongyi Deep Research, the Leading Open-source Deep Research Agent
概览
--- name: arbor description: 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. allowed-tools: Read Write Edit Bash Agent license: MIT license metadata: version: "1.1" skill-author: K-Dense Inc. ---
# Arbor — Autonomous Optimization via Hypothesis Tree Refinement
## Overview
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.
It implements **Hypothesis Tree Refinement (HTR)** from *Arbor* (Jin et al., 2026). The key idea: keep the research state in a persistent **hypothesis tree** rather than in conversation history. Each node binds a hypothesis, the distilled insight it produced, and a pointer to the artifact version that realizes it. You play the long-lived **coordinator** that owns this tree and decides where to search; short-lived **executor** subagents test one hypothesis each in isolated git worktrees and report back. A **held-out merge gate** admits a change only when it improves on a *test* evaluator the search never optimized against. This is what turns trial-and-error into cumulative, auditable research.
Use the `scripts/tree.py` state manager for all the bookkeeping (creating nodes, writing evidence, propagating insights, pruning, the merge gate, the Observe projection). It keeps the state consistent and frees you to spend judgment on what the evidence *means*.
## When to use this skill
Reach for Arbor when the task is **iterative improvement of a concrete artifact under an evaluator**: - Model training: optimizer/architecture/recipe changes to lower loss or hit a target in fewer steps. - Harness/agent engineering: raising pass rate or accuracy of an agent loop, search harness, or tool-use scaffold. - Data synthesis: improving a generation/filtering pipeline judged by downstream model behavior. - Benchmark optimization: MLE-bench / Kaggle-style "improve the submission" tasks. - Prompt/system optimization where you can score outputs automatically.
The distinguishing signals: there's an **artifact you can modify**, an **objective**, a way to **score** candidates, and you expect to run **many experiments**. If the user only wants a single fix or a one-shot answer, this is overkill — just do the work directly. If they want open-ended ideation with no evaluator, use `hypothesis-generation` or `scientific-brainstorming` instead.
## The AO setup — pin this down first
Before any experiments, establish the task tuple `(M_0, O, E_dev, E_test)`. Getting this right matters more than any later decision, so confirm it explicitly:
- **M_0 — initial material**: the artifact to improve (a repo, a script, a config, a prompt). Make sure it's under git and currently runs. - **O — objective**: the natural-language goal and the metric *direction* (maximize accuracy? minimize loss/steps?). - **E_dev — development evaluator**: a command you can run freely during search to score a candidate. Fast, repeatable. - **E_test — held-out test evaluator**: a *separate* evaluator (different seeds, different split, or a larger run) used only at the merge gate. It must not be used as a search oracle — that's the whole point.
If the user hasn't given you a clean dev/test split, **construct one and say so**. The dev/test separation is the mechanism that catches overfitting: a candidate that wins on dev but not on test isn't a success, it's a warning that you're exploiting the feedback signal. Without it, autonomous search reliably overfits.
Initialize the run:
```bash python scripts/tree.py init \ --objective "Improve BrowseComp answer accuracy on the search harness" \ --dev-eval "python eval.py --split dev --n 50" \ --test-eval "python eval.py --split test --n 300" \ --material "." --metric-direction max --branching 3 --max-depth 2 --budget 12 ```
`--branching` is how many sibling hypotheses you propose per parent; `--max-depth 2` keeps directions at depth 1 and concrete interventions at depth 2 (the paper's default); `--budget` is the number of coordinator cycles. Start small (10–20 cycles) — structured search beats brute force, and you can extend if progress is still being made.
## The coordinator loop
You run repeated cycles of six steps. This is the heart of HTR; do not collapse it into ad-hoc editing. Run `python scripts/tree.py cycle` once per cycle to track the budget.
### 1. Observe Begin every cycle by re-grounding in the tree, not in your memory of the conversation:
```bash python scripts/tree.py observe ```
This prints the objective, global insights, the active frontier (selectable hypotheses), executed nodes with their evidence, pruned lessons (negative constraints), and the current best artifact. Treating the tree as the source of truth is what keeps you coherent over a long run, after context compression has thrown away the details.
### 2. Ideate Pick a promising parent and propose a few child hypotheses under it. **Condition on the tree's evidence** — this is the difference between Arbor and random search: - Validated insights are assumptions you can build on. - Pruned nodes are dead ends to avoid. - A "half-right" result is a *starting point for a sharper hypothesis*, not a reason to abandon the direction.
Each hypothesis should be a **falsifiable claim about how changing the artifact will move the metric**, not a vague intention. Depth-1 nodes are broad directions ("the search harness loses correct answers it already retrieved"); depth-2 nodes are concrete, executable interventions ("run K=5 independent rollouts and aggregate by evidence dossier instead of majority vote").
```bash python scripts/tree.py add-node --parent n0 --hypothesis "Verification, not retrieval, is the bottleneck: candidates are found but discarded" python scripts/tree.py add-node --parent n4 --hypothesis "Decompose the question into atomic constraints and verify each independently" ```
### 3. Select Choose which pending leaves to run next. **Selection is not pure score-maximization** — pick a hypothesis because it has strong prior evidence, because it would resolve an ambiguity its siblings exposed, or because its failure would clarify an important assumption. Frontier control under delayed feedback rewards informative experiments, not just promising ones.
### 4. Dispatch Run each selected hypothesis as an **executor subagent in an isolated worktree** (use the Agent tool with `isolation: "worktree"`, or have the executor create one with `git worktree add`). Isolation matters: parallel experiments must not clobber each other or the current best, and exploratory changes stay quarantined until they pass the merge gate.
Dispatch siblings **in parallel** (multiple Agent calls in one message) when they're independent — comparative evidence within one direction is exactly what makes later pruning and abstraction possible.
Give each executor a tight, **hypothesis-bound** brief. See `references/executor-brief.md` for the full template. The contract that makes HTR work: **the executor may not change the hypothesis when the metric stalls.** It repairs its own code and reruns, but `h_n` is fixed — otherwise the returned score is no longer evidence about the assigned node and the tree's semantics break. The executor returns exactly four things: - **dev_score** — the dev evaluator result (for selection); - **result** — a factual summary of what happened; - **insight** — the distilled, reusable lesson (*why* the result supports, weakens, or bounds the hypothesis); - **branch_ref** — the git branch/commit/worktree path holding the artifact.
Mark a node `running` before dispatch (`tree.py set-status --node n5 --status running`) so the Observe projection stays accurate.
### 5. Backpropagate When an executor returns, write its report into the node, then **abstract the lesson upward**:
```bash python scripts/tree.py set-evidence --node n5 --dev-score 70.0 \ --result "K=5 dossier aggregation recovers answers in minority rollouts" \ --insight "Correct answers often appear in a minority of rollouts; aggregation beats majority vote" \ --branch-ref "wt/n5"
python scripts/tree.py propagate --node n5 \ --insight "Candidate coverage, not verification, limits this direction" --to-root ```
This is the step that makes the tree more than a log. A leaf-level observation ("data-interface mismatch") should become a direction-level constraint and, if it generalizes, a global prior that shapes future ideation. **Insight propagation is the component that drives most of HTR's gains** — in the paper's MLE-Bench Lite ablation, a tree *without* insight feedback scored even lower than a flat experiment queue with no tree at all (54.5% vs. 63.6% any-medal, against 81.8% for the full system). Hierarchy alone isn't enough: the semantic memory is what matters. So spend real thought on the abstraction; don't just copy the leaf insight upward verbatim.
### 6. Decide Decide what to do with the new evidence: keep expanding a direction, prune a falsified subtree, or attempt to merge a candidate.
- **Prune** dead ends, recording *why* — the reason becomes a negative constraint: ```bash python scripts/tree.py prune --node n7 --reason "search-augmented judge overfits dev questions; no test transfer" ``` - **Merge gate** — promote a candidate to the new best **only if it improves on `E_test`**. Run the test evaluator in a *fresh* worktree (not the dev worktree, to avoid leakage), then: ```bash python scripts/tree.py merge --node n5 --test-score 67.67 --branch-ref "wt/n5" ``` If the gate rejects it, that's informative: a high-dev / low-test candidate is evidence the direction may be exploiting the dev signal rather than producing a transferable improvement. Record that lesson; don't quietly promote it anyway.
Repeat until the budget is spent, the frontier is exhausted, or progress has clearly stalled.
## Finishing the run
When you stop, produce a short report (see `references/report-template.md`) covering: - the final best artifact, its test score, and its delta over `M_0`; - the tree (`python scripts/tree.py status`) as the audit trail of what was tried; - the main hypothesis shifts — how task understanding deepened across the run (early nodes test broad mechanisms; later nodes find their limits; ancestor insights compress these into the constraints behind the final design); - merged vs. explored: many nodes improve dev, far fewer pass the test gate — report that gap honestly rather than overstating dev wins.
Always leave `M_best` as a real, runnable artifact on a named branch, and tell the user how to check it out.
## Principles that make this work (not rote rules)
These come from the paper's analysis; understanding *why* matters more than following them mechanically.
- **The tree is the memory; conversatio
技术详情
- 版本
- 1.0.0
- 许可证
- MIT license
- 最近更新
- 2026年8月20日
- 发布时间
- 2026年8月20日
决策摘要
首选
33,974 个 GitHub Stars
Agent 验证证据
Agent 验证证据
来自解析、审查、安装和一次小范围运行后的结果报告。
- 成功率
- —
- 近期失败
- —
- 结果
- 0
- 输出质量
- —
- 失败
- 0
- 不相关
- 0
- 安装次数
- 0
- 风险拦截
- 0
- 需要配置
- 0
- 生产环境
- 0
暂时没有 Agent 结果数据。首次 Agent 执行可以通过 /api/agent/outcome 报告成功、需要设置、风险拦截、失败或不相关。
增长闭环
分享工具包
为 arbor 准备的场景化草稿,可手动发布到 X。
A practical pick for source-backed research: arbor: Autonomously improve a real artifact (code, training recipe, agent harness, data pipeline, prompt) against an objective and... 34.0K stars https://www.openagentskill.com/skills/k-dense-ai-arbor?ref=x
可选:带安装命令的回复
Listing + install path for arbor: https://www.openagentskill.com/skills/k-dense-ai-arbor?ref=x Install: npx skills add K-Dense-AI/scientific-agent-skills --skill arbor
收录来源
Registry 收录
此列表来自公开来源,维护者认领获批前不会标记为官方。
- 创作者
- K-Dense-AI
- 收录方
- OpenAgentSkill 社区索引
归属链接指向公开仓库或创作者主页。创作者可认领列表以更新所有权信号。
认领此 Skill所有者认领
认领此 Skill 页面
这条 Registry 收录 列表归属于 K-Dense-AI,但尚未标记为官方。认领后可增加已验证所有者信号,使后续发布、安装和审计更新更值得信赖。
创作者外链工具包
将证据徽章加入你的 README
在开发者评估仓库的位置展示规范页面、当前信任与审计信号,以及真实的 Agent 验证证据。
[](https://www.openagentskill.com/skills/k-dense-ai-arbor)
[](https://www.openagentskill.com/skills/k-dense-ai-arbor)
[](https://www.openagentskill.com/skills/k-dense-ai-arbor/audit)
[](https://www.openagentskill.com/skills/k-dense-ai-arbor)作者
K-Dense-AI
@k-dense-ai
平台适配
健康信号
- GitHub Stars
- 34.0K
- 质量评分
- 55/100
- 最近 GitHub 推送
- 2026年8月20日
- 框架提示
- 未知
- OpenAgentSkill 浏览量
- 19
- 复制安装命令
- 0
- 跳转点击
- 0
社区信号
告诉我们这个 Skill 是否对你的 Agent 工作流有帮助。汇总反馈会持续改善排序。
信任与安全
审查后安装
- GitHub 采用度34K 个 GitHub Stars通过
- Star/Fork 活跃度34K 个 Star,3.3K 个 Fork; 当前元数据中没有议题活跃度信息通过
- 近期维护距上次推送 2 天通过
- 许可证清晰度MIT license通过
- README/SKILL.md 完整度公开元数据需要更完整的 README/SKILL.md 上下文信息
- 依赖与运行时风险命令执行范围信息
相关 Skill
Last30days Skill
Research the last 30 days across Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web, then synthesize a grounded brief for an AI agent.
53.5K StarsAcademic Research Skills
Academic Research Skills for Claude Code: research → write → review → revise → finalize
38.4K StarsGPT Researcher
Run autonomous deep research over web and local sources
28.0K StarsDeepResearch
Tongyi Deep Research, the Leading Open-source Deep Research Agent
19.8K Stars