← 返回 Skills 市场
Meta Research
作者
Jiachen LIU
· GitHub ↗
· v2.5.0
569
总下载
0
收藏
1
当前安装
7
版本数
在 OpenClaw 中安装
/install meta-research
功能描述
Autonomous research workflow agent for AI and scientific research. Use when the user wants to brainstorm research ideas, conduct a literature review, design...
安全使用建议
This skill is internally consistent with its stated purpose and appears to be a well-documented research workflow. Before installing or enabling autonomous runs: (1) Be aware the agent will create and modify files in your project (LOGBOX, explorations/, shared/). Back up any important data first. (2) The skill can run shell commands and make web requests — avoid granting persistent credentials or broad network access unless you trust the skill and monitor actions. (3) If you plan to publish or push code/data, prepare and control any GitHub/Zenodo/API credentials separately and provide them only when needed. (4) Consider running the skill in a sandboxed project folder or ephemeral environment until you confirm its behavior. If you want deeper assurance, request the author/source/repository before installing so you can audit exact file changes and any optional scripts.
功能分析
Type: OpenClaw Skill
Name: meta-research
Version: 2.5.0
The skill is classified as suspicious due to its broad `allowed-tools` permissions, specifically `Bash` and extensive file system access (`Read`, `Write`, `Edit`, `Glob`, `Grep`). While the skill's instructions across all `.md` files are consistently aligned with its stated purpose of an 'Autonomous research workflow agent' and promote rigorous, reproducible, and ethical research practices, the inherent capability to execute arbitrary shell commands (`Bash`) and manipulate the file system presents a significant vulnerability for potential remote code execution or unauthorized file operations if the agent's prompt handling or execution environment were compromised. There is no evidence of intentional malicious behavior such as data exfiltration, backdoor installation, or obfuscation within the provided files.
能力评估
Purpose & Capability
The name/description (autonomous research copilot) matches what the skill asks for: structured phase protocols, writing files, literature search, and reproducible experiment management. There are no unrelated required environment variables, binaries, or install steps that would be inconsistent with the described purpose.
Instruction Scope
SKILL.md instructs the agent to read and write project files (LOGBOX, explorations/*, shared/*), create directories, run literature searches (arXiv, Semantic Scholar, Google Scholar), and manage reproducibility artifacts. Those actions are appropriate for a research workflow. It does not instruct reading arbitrary system files or requiring unrelated secrets, but it does assume the agent can perform file I/O, shell commands (Bash) and web requests — capabilities that can perform broader actions if granted. The instructions are specific (phase protocols, artifact locations) rather than open-ended data collection.
Install Mechanism
This is an instruction-only skill with no install spec and no code files to execute on install. That is the lowest-risk install model and proportionate to the stated functionality.
Credentials
No environment variables, credentials, or config paths are declared or required. The workflow mentions interacting with services (GitHub, arXiv, Zenodo) as part of dissemination, but the skill does not request tokens or secrets in its metadata. If the agent later asks for upload credentials at runtime, those are not declared here and should be provided explicitly by the user only when needed.
Persistence & Privilege
always:false and the skill does not request system-wide persistence. The skill is allowed to run autonomously by default (platform default), and it lists powerful allowed-tools (Bash, WebFetch, WebSearch, file Read/Write/Edit). That is coherent for an autonomous research helper but increases the blast radius if you enable autonomous execution — consider restricting autonomous runs or reviewing prompts before allowing shell/network actions.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install meta-research - 安装完成后,直接呼叫该 Skill 的名称或使用
/meta-research触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v2.5.0
Latest sync from GitHub
v2.2.0
Sync with latest GitHub
v2.1.0
Latest from GitHub
v2.0.0
v2.0: hypothesis-driven workflow, research-tree.yaml, judgment gate, reflection phase, hypothesis-generation and experiment-execution phases
v1.0.2
Add marketplace.json, ideation frameworks F1-F9, explorations/ structure, templates
v1.0.1
Add ideation-frameworks.md (F1-F9) and fix file structure
v1.0.0
Initial release: full research lifecycle skill with brainstorming frameworks (F1-F9), lit review, experiment design, analysis, and writing phases. Includes LOGBOX tracking, FINER scoring rubric, reproducibility checklist, and ideation frameworks.
元数据
常见问题
Meta Research 是什么?
Autonomous research workflow agent for AI and scientific research. Use when the user wants to brainstorm research ideas, conduct a literature review, design... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 569 次。
如何安装 Meta Research?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install meta-research」即可一键安装,无需额外配置。
Meta Research 是免费的吗?
是的,Meta Research 完全免费(开源免费),可自由下载、安装和使用。
Meta Research 支持哪些平台?
Meta Research 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 Meta Research?
由 Jiachen LIU(@amberljc)开发并维护,当前版本 v2.5.0。
推荐 Skills