WorkBuddy + Academic-Research-Skills:保姆级图文教程WorkBuddy + Academic-Research-Skills: A Step-by-Step Guide with Pictures


AI free token/AI免费积分

本文来源:https://mp.weixin.qq.com/s/uh66xy8BulhA01YLGNDtnA
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

前几篇讲 Codex 的教程发出去,后台收到最多的留言是两个:一是科学上网太难,二是想干点学术活,但 Codex 那套门槛还是高了点。

WorkBuddy 刚好把这两个问题一起解了。

它是腾讯团队做的 AI 工作台,不用科学上网,手机号注册就能用,最近热度一直往上走,身边不写代码的朋友也在用。

这篇讲怎么把 academic-research-skills(下面简称 ARS)这套学术研究技能装进 WorkBuddy,让它在你这个 AI 工作台里多一条专业的研究线。

先把一句容易误会的话讲在前面。

图片

ARS 不是替你写论文的工具。 它帮你查文献、排版引用、核对数据、检查逻辑,但定义研究问题、选方法、解读数据这些得你自己来。

官方 README 反复强调一句话——「副驾驶,不是代驾」。

照着这篇一步步做,没用过 WorkBuddy 也能装通。


先分清两个容易混的概念:技能 vs 连接器

装之前先把概念讲清楚,不然后面你会懵。

WorkBuddy 有两套扩展机制,名字像,干的事完全不一样:

  • 技能(Skill):回答「AI 能帮我做什么新活」。给 AI 一套新的工作能力。ARS 就是技能。
  • 连接器(Connector):回答「AI 能连到我的哪些数据」。让 AI 能读写你授权的外部账号(飞书、Notion、GitHub、腾讯文档)。
图片

图 1:技能和连接器是两套不同的扩展机制

说白了就是一句话:技能管「能力」,连接器管「数据来源」。

两者会配合——某个技能要读你的腾讯文档时,才需要先连好腾讯文档这个连接器。

但这篇只装技能(ARS),不碰连接器,先把主线走通。


第一步:把 WorkBuddy 装上

已经装了 WorkBuddy 的,跳过这步。

没装的话,打开官网 workbuddy.cn,首页有个「下载 WorkBuddy」按钮,点一下是什么操作系统,就下什么安装包,一路回车即可。

图片

打开后用手机号注册登录,进主界面就行。这步没什么坑。

✅ 到这里你应该已经:能正常打开 WorkBuddy,登录进去看到主界面。

图片

第二步:找到「技能」入口,选一种方式装 ARS

先在 WorkBuddy 里找到「技能」在哪,再选下面两种方式之一把 ARS 装上。

两种方式殊途同归,差别在版本新旧和操作门槛,后面对比讲。

登录后,左侧那列导航里找「技能」这个标签,点进去。

图片

你会看到两个区域:

  • 技能市场:官方推荐的技能,能一键装
  • 已安装:你本地装好的技能,ARS 装完会出现在这里
图片
图片

方式 A:技能市场一键安装(最简单)

适合不想折腾命令行、只想先跑通流程的人。

在「技能市场」里找到 ARS 的对应技能(名字通常是 academic-research-skills 或拆出来的几个子技能),点一键安装即可。

图片

注意:技能市场里的版本是 WorkBuddy 平台打包上架的,更新节奏慢半拍——作者在 GitHub 刚发的新功能、bug 修复,可能要等一段时间才同步到市场。如果你只是想先试用、不追新,方式 A 完全够用。

装完直接跳到第四步验证就行,不用走第三步的打包上传。

方式 B:从 GitHub 拉源码,打包上传(拿最新版)

ARS 开源在 GitHub:Imbad0202/academic-research-skills。

想用最新版、或者技能市场里还没上架对应技能时,就自己从仓库拉源码、打包成技能包、再上传。这条链路完整覆盖作者最新提交。

打开终端,把仓库克隆到本地:

# 克隆 ARS 仓库到主目录(需要先装 git)
cd ~
git clone https://github.com/Imbad0202/academic-research-skills.git

💡 没装 git? 两个办法。一是终端运行 xcode-select --install 装 Xcode 命令行工具,自带 git。二是直接去仓库页面点绿色「Code」按钮 → 「Download ZIP」,下完解压,解压出来的文件夹就当克隆的仓库用,后面路径换一下。

拉完验证一下,进仓库目录看看有没有四个子技能:

cd ~/academic-research-skills
ls

列表里能看到 deep-research、academic-paper、academic-paper-reviewer、academic-pipeline 这四个目录,就对了(还有别的文件,不用管)。

✅ 检查点:ls 能看到这四个目录名。

两种方式怎么选

到这里你应该已经明白差别了,简单收一下:

维度
方式 A:技能市场
方式 B:GitHub 源码
版本新鲜度
慢半拍,平台打包上架后才同步
最新
,跟着作者提交走
操作门槛
点一下就行,零命令行
要装 git、敲命令、打包、手动上传
稳定性
平台审核过,兼容性有保障
偶尔碰到作者刚提交的 bug
适合谁
先试用、不追新的人
想要新功能、愿意折腾的人

一句话:想省事走 A,想要新功能走 B。 走 A 的装完直接跳第四步验证;走 B 的继续往下,把源码打包传进 WorkBuddy。


第三步:打包,再传进 WorkBuddy(仅方式 B)

这步只针对选了方式 B 的人,说清楚就一件事——把四个子技能各自压成一个 zip,传到 WorkBuddy 的「技能」里。

3.1 认识 ARS 的四个子技能

先知道这四个分别是干嘛的,不然装完不知道用哪个:

图片

图 2:ARS 的四个子技能,覆盖从选题到完整论文

拆开看其实很简单:

  1. deep-research:搞清楚「研究什么、怎么研究」。做文献综述、系统性回顾、事实查核,也能用提问式对话帮你厘清模糊的研究方向。新手从这个开始。
  2. academic-paper:把研究写成规范论文。支持多种论文类型和引用格式,能出 LaTeX / Word / PDF。
  3. academic-paper-reviewer:投稿前模拟一轮同行评审,多个虚拟审稿人视角挑刺。
  4. academic-pipeline:把上面三个串成完整闭环。功能最全,但消耗大、耗时长,新手别上来就用。

本篇第四步会用门槛最低的 deep-research 跑第一个任务。

3.2 打包成技能包

WorkBuddy 导入技能有个硬要求:**zip 包的顶层必须直接是 SKILL.md**,不能外面再套一层文件夹。

ARS 官方文档正好给了打包命令,直接用。在终端、确保你还在 ~/academic-research-skills 目录下,执行:

# 四个子技能分别打包成 zip
# 每个包的顶层直接是 SKILL.md,符合 WorkBuddy 的技能包格式
for s in deep-research academic-paper academic-paper-reviewer academic-pipeline; do
  (cd"$s" && zip -rq "../$s.zip" . -x "*.DS_Store")
done

-q 是静默模式,没输出是正常的。

验证四个包都生成了:

ls -lh *.zip

能看到这四个文件,大小在 100K–350K,就对了:

academic-paper-reviewer.zip
academic-paper.zip
academic-pipeline.zip
deep-research.zip

💡 怎么确认格式没套错:Finder 里双击某个 zip 解压,解压出来的文件夹里应该直接看到 SKILL.md。如果解压出来先是一层「deep-research」文件夹、点进去才是 SKILL.md,那就是套错了——回到上面命令,确认是 cd 进子技能目录、用 zip -r "../xxx.zip" . 打包当前内容,不是把整个文件夹拖进去压。

✅ 检查点:ls *.zip 能列出这四个包,大小都在几十 KB 以上。

3.3 传进 WorkBuddy

回到 WorkBuddy,左侧「技能」→ 找「添加技能」按钮 → 选「上传技能」。

图片

选「上传技能」后会出现一个文件选择框或可拖拽区域。

把 deep-research.zip拖进去(或点「选择文件」选中它)。WorkBuddy 会自动解析配置,传完这个技能就出现在「已安装」里了。

然后重复同样的操作,把剩下三个也依次传上去:academic-paper.zip、academic-paper-reviewer.zip、academic-pipeline.zip。

图片

💡 为什么不能一次传四个:WorkBuddy 的上传是「一个 zip 对应一个技能」,每个 zip 顶层只能有一个 SKILL.md。所以得分四次传,每次选一个。

四个全传完,「已安装」里应该能看到这四个名字。

✅ 到这里你应该已经:在「技能 → 已安装」里看到 ARS 的四个子技能,都是启用状态。


第四步:跑通第一个任务,确认它真在工作

装完不算完,得真跑一次,确认技能被调用了。

别上来就用 academic-pipeline

四个里,先用 deep-research。原因很实在:

  • 不需要你先准备论文素材
  • 有个「苏格拉底模式」,你说一句模糊的研究兴趣,它反过来提问帮你厘清
  • 积分消耗相对小,适合第一次试水

academic-pipeline 那种端到端的全流程,单次积分消耗会明显高于普通对话。

体验版每月 500 积分,跑一两次重度任务可能就见底了。所以入门先用 deep-research 试水。

跑一次选题探索

点左侧「新建任务」,在对话框里用大白话描述你的研究兴趣。比如:

我对「AI 工具对高校教学评估的影响」这个方向有点兴趣,
但还没想清楚具体研究什么,能帮我理一理吗?

发出去之后,看它怎么回。

判断 ARS 被调用的信号:它不会直接甩给你一段答案,而是反过来问你——

「你关注的是哪个学段?是评价教学质量还是学生学习效果?有没有想聚焦的国家或地区?」

这就是 deep-research 的苏格拉底模式在工作。

图片
图片

如果它直接长篇大论给一段「分析」,没提问、没说要调技能——可能是技能没被自动识别,看下面的排错。

它问,你答,聊几个来回。

通常 5–15 轮,你会得到一个聚焦的研究问题和方法方向。觉得差不多了,让它把当前方向整理成一段小结。

到这里,安装和验证就闭环了。

✅ 到这里你应该已经:用 deep-research 完成一次研究方向探索,亲眼看到技能被调用、在工作。


平时怎么管:开、关、卸,还有积分

装好之后,几件要知道的事。

只开当前任务需要的技能。 「技能 → 已安装」里每个技能都能单独开关,不用卸载。

一次性开太多,AI 容易在该用 A 的时候误触发 B,反而降低准确率。平时用 deep-research 就先只开它,要写论文再开 academic-paper。

积分消耗心里有数。 ARS 这种技能,一个技能背后是十几个智能体在协作,跑一次完整任务消耗通常比普通对话高,尤其 academic-pipeline。

要跑重度任务前,先去「账号 → 用量查看」确认剩余积分够不够,不够就补套餐或等次月刷新。

确定不用了再卸。 在「已安装」里选「卸载」,支持批量。卸载后从本地删除,需要时重新传 zip 装回来。


常见坑

点「上传技能」后导入失败。 十有八九是 zip 格式不对——顶层不是直接的 SKILL.md,而是套了一层文件夹。

双击 zip 解压看一眼就知道了。解法:回到第三步,确认是 cd 进子技能目录打包当前内容,重新压。

技能装好了,对话时没被调用。 先看「技能 → 已安装」里目标技能是不是启用状态;再把其他技能关掉、只留当前需要的,重发一次。还不行重启 WorkBuddy。

技能调用了,跑到一半报错或中断。 多半是积分不足或网络波动。到「账号 → 用量查看」看积分,网络问题直接重试。

macOS 提示「无法打开 WorkBuddy」。 回到第一步的处理:先删掉重新下;还不行在「系统设置 → 隐私与安全性」找「仍要打开」。


再说一句

WorkBuddy 加上 ARS,等于把「查文献、做综述、写论文」这套专业流程,接进了一个能真正干活的 AI 工作台。

装技能这件事本身不难,难点始终在你怎么定义研究问题、怎么判断 AI 给的结果靠不靠谱——这部分,工具帮不了你,也正是「不可替代」的那部分。

工具能帮你省的是重复劳动,不能帮你省的是判断。


参考来源

  • WorkBuddy 官网 https://www.workbuddy.cn/
  • WorkBuddy 文档站 https://www.workbuddy.cn/docs/workbuddy/From-
  • academic-research-skills https://github.com/Imbad0202/academic-research-skills

 

Source: https://mp.weixin.qq.com/s/uh66xy8BulhA01YLGNDtnA If there are any copyright issues, please contact: 13951431913#139.com or submit a comment on the right.

The most common messages I receive after posting tutorials on Codex are: one, setting up a VPN is too difficult; two, I want to do some academic work, but the threshold for using Codex is still quite high.

WorkBuddy has come in to solve both of these problems.

It is an AI workbench developed by Tencent. You do not need a VPN to use it; you can sign up with just your phone number. It has been gaining popularity recently, with even non-coders in my circle using it.

This tutorial explains how to install the academic-research-skills (ARS) into WorkBuddy, adding a professional research capability to your AI workbench.

Before we proceed, I must clarify one thing.

图片

The official README emphasizes this point repeatedly: “I am your copilot, not your driver.”

Following the steps in this guide, even those who have never used WorkBuddy before should be able to set this up without issue.

Step One: Install WorkBuddy

If you have already installed WorkBuddy, you can skip this step.

If not, go to the official website at workbuddy.cn. Click the “Download WorkBuddy” button on the homepage. The site will prompt you for your operating system; download the corresponding installer and follow the on-screen instructions to install it.

Once the installation is complete, open the application and log in using your phone number. You should now be in the main interface.

✅ At this point, you should: be able to open WorkBuddy and log in to the main interface.

Step Two: Locate the “Skills” Entrance and Install ARS

First, find where the “Skills” section is located in WorkBuddy, then choose one of two methods to install ARS.

These two methods achieve the same goal but differ in terms of software version and operational complexity, which we will compare later.

After logging in, look for the “Skills” tab in the left-hand navigation menu and click on it.

You will see two areas:

Skill Market: This contains official recommended skills that can be installed with a single click.

Installed: This lists the skills that have already been installed on your local machine. ARS will appear here once it is successfully installed.

Option A: One-Click Installation via the Skill Market (Simplest)

This method is ideal for those who do not want to deal with command lines and simply want to get the tool running as quickly as possible.

In the “Skill Market”, locate the entry for ARS (usually named “academic-research-skills” or its sub-components) and click “Install”.

Step 2: Locate the “Skills” entry point and select a method to install ARS.

Once installed, you can proceed directly to Step Four to verify that the skill is working. There is no need to go through the packaging and uploading process in Step Three.

Option B: Download Source Code from GitHub, Package, and Upload (For the Latest Version)

ARS is an open-source project hosted on GitHub at Imbad0202/academic-research-skills.

If you need the latest version or wish to use features that haven’t been added to the Skill Market yet, you can manually download the source code, package it into a skill file, and upload it to WorkBuddy. This ensures you are working with the most up-to-date version straight from the developer.

Open your terminal and clone the repository to your local machine:

If you don’t have git installed, there are two solutions: you can install the Xcode command-line tools (which includes git) by running xcode-select --install in the terminal, or you can navigate to the GitHub page and download the source as a ZIP file, extracting it for use later.

Once the repository is cloned, navigate into the directory to verify its contents:

You should see four sub-directories: deep-research, academic-paper, academic-paper-reviewer, and academic-pipeline. These are the individual components of the ARS toolkit.

✅ At this point, your terminal should output the names of these four directories in the list.

How to Choose Between the Two Options

By now, the differences between the two methods should be clear. Here is a quick summary:

Criterion

Option A: Skill Market

Option B: GitHub Source

Version Timeliness

Slightly delayed; depends on platform review

Immediate; directly from the author

Installation Difficulty

Very easy; one-click

More complex; requires git, command line, and manual upload

Stability

High; vetted by the platform

Varies; depends on the current state of the source code

Recommended For

Beginners, casual users

Advanced users, those requiring specific new features

In short, choose A for convenience and B for cutting-edge features.

Step Three: Package and Upload to WorkBuddy (Only for Option B)

This step is necessary only if you have chosen to install ARS from the source code (Option B). The goal is to compress each of the four sub-modules into a ZIP file and upload them to WorkBuddy.

3.1 Understanding the Four ARS Modules

Before we package the files, let's briefly go over what each of these modules does, so you know what to expect when you start using them:

deep-research: This is for figuring out what to research and how to do it. It assists with literature reviews, systematic reviews, fact-checking, and clarifying ambiguous research directions. It is a good starting point for beginners.

academic-paper: This module is for writing the actual paper. It supports various paper types and citation formats, and it can export LaTeX, Word, and PDF documents.

academic-paper-reviewer: This allows you to simulate peer review. It generates feedback from multiple virtual reviewers to help you improve your draft before submission.

academic-pipeline: This is the "kitchen sink" module that combines all the others into a single, comprehensive workflow. It is very powerful but also the most computationally expensive, so it is best reserved for later stages.

In this tutorial, we will start by testing the deep-research module in Step Four.

3.2 Packaging the Modules

WorkBuddy requires that the top level of the uploaded ZIP file is the SKILL.md document. There cannot be an extra folder layer at the top.

Fortunately, the official ARS documentation provides a command to do this correctly. In your terminal, while still inside the ~/academic-research-skills directory, run the following commands:

The -q flag is for quiet mode, meaning you won't see a lot of output, which is expected if the command runs successfully.

Once the command finishes, verify that the four ZIP files have been created:

You should see the four files listed, each roughly between 100 KB and 350 KB in size.

To double-check the formatting, unzip one of the files and look at the contents. The extracted folder should contain a file named SKILL.md immediately inside, rather than a sub-folder containing that file. If you see a sub-folder, the packing command was likely run from the wrong directory. Re-run the command from the root of the repository if necessary.

✅ At this point, you should have four ready-to-upload ZIP files.

3.3 Uploading to WorkBuddy

Return to the WorkBuddy application.

Go to the "Skills" section in the left navigation bar and click the "Add Skill" button.

Select "Upload Skill". A file selector dialog will open.

Upload the first file, deep-research.zip. WorkBuddy will automatically parse the configuration, and the skill will appear in your "Installed" list once the upload is complete.

Repeat the process for the remaining three files: academic-paper.zip, academic-paper-reviewer.zip, and academic-pipeline.zip.

It is important to note that WorkBuddy requires you to upload these files one at a time as single ZIP archives. You cannot upload a folder of multiple ZIPs simultaneously.

✅ By the end of this step, you should see all four ARS modules listed in the "Installed" section of the Skills menu.

Step Four: Running Your First Task to Verify Functionality

Installing the software is only half the battle; you need to verify that it actually works by running a task.

For your first test, use the deep-research module.

Avoid using the academic-pipeline module for your initial test. While it is very powerful, it is also the most complex and computationally intensive. Using it immediately might consume a significant portion of your monthly quota before you have even verified that the setup is correct. By starting with deep-research, you can confirm that the installation works without risking a large portion of your credit balance.

To begin, click the "New Task" button in the left navigation bar.

In the dialogue box that appears, describe your research interest in plain language. For example:

"I want to study how online education platforms impact student learning outcomes in Asia."

Send the message and observe the response.

A successful interaction with the ARS skill will result in the AI asking you follow-up questions rather than providing a direct, static answer. This "Socratic" approach is the hallmark of the deep-research module.

If the AI responds with a long, monologue-style answer immediately, it suggests that the skill is not being invoked correctly or that the system is defaulting to a general model rather than the specialized research assistant.

If the interaction goes well, continue the conversation for 5 to 15 rounds, answering the questions posed by the AI. At the end, ask the AI to summarize the current research direction, which should consolidate your discussions into a coherent plan.

✅ By the end of this step, you should have completed a session where the deep-research module actively engaged with you to refine a research topic, confirming that the installation is fully functional.

Managing Your Skills and Credits

Now that the system is up and running, there are a few housekeeping items to keep in mind:

Selective Activation: In the "Installed" menu, you can toggle each skill on or off. It is generally a good practice to only activate the skills you intend to use for a specific task. Having multiple, unrelated skills active simultaneously can sometimes confuse the AI or slow down the response time unnecessarily.

Credit Awareness: As mentioned earlier, skills like academic-pipeline consume a lot of computing power and thus a lot of credits. It is advisable to check your remaining balance in the "Account" section before attempting to run these more extensive tasks. If your credits are low, you may need to purchase more or wait for the monthly refresh.

Uninstallation: If you find that a skill is not working as expected or you no longer need it, you can delete it from the "Installed" list. Deleting the skill removes it from your local configuration, but you can always re-upload the ZIP file to install it again later.

Common Troubleshooting Scenarios

Import Failure: If the "Upload Skill" button fails to import the ZIP file, the most likely cause is incorrect formatting. Specifically, the ZIP file must contain the SKILL.md file at the root level. If the ZIP contains a folder that contains the SKILL.md file, the import will fail. Unzip the file and verify its structure before uploading.

Skill Not Invoked: If the AI seems to be "blind" to the skill, i.e., it gives generic responses instead of the specialized dialogue expected from the ARS, check that the skill is actually toggled "On" in the management panel. Also, ensure that no other conflicting skills are activated that might be interfering with the prompt. A full restart of the WorkBuddy application can also resolve initialization errors.

Task Interruption: If a task starts but crashes mid-way, this is often attributed to insufficient credits or a temporary network glitch. Check your credit status in the account dashboard and try re-running the task. If it is a network issue, simply retrying the request should suffice.

macOS Blocking the App: If macOS prevents the application from opening, delete the existing installation and download it again from the official website. If the prompt to allow the app persists, navigate to "System Settings" > "Privacy & Security" and find the "Allow" option for WorkBuddy.

Final Thoughts

Integrating the academic-research-skills into WorkBuddy effectively transforms a general-purpose chatbot into a specialized research assistant. This setup automates the mundane but critical tasks of literature review and citation formatting, allowing researchers to focus their mental energy on the creative aspects of their work, such as hypothesis generation and data interpretation.

While the installation process requires some initial setup, the long-term benefits of having this functionality available without the need for a VPN or complex cloud configurations are significant. As with any tool, the quality of the output is directly dependent on the quality of the input; the tool can do the heavy lifting, but it cannot replace the analytical judgement of the user.

Sources

WorkBuddy Official Website: https://www.workbuddy.cn/

WorkBuddy Documentation: https://www.workbuddy.cn/docs/workbuddy/From-

academic-research-skills GitHub Repository: https://github.com/Imbad0202/academic-research-skills

图片

图片

图片

图片

图片

图片

图片

图片

图片

图片

图片

图片

图片

Combining WorkBuddy with ARS effectively integrates the professional workflow of “literature retrieval, review drafting, and paper writing” into an AI-powered work platform capable of real execution.

While installing the skill is mechanically simple, the core challenge remains in how you formulate your research questions and evaluate the credibility of the AI-generated results — areas where the tool cannot substitute for human expertise.

The tool helps eliminate repetitive tasks but cannot replace critical judgment.

参考来源

WorkBuddy 官网 https://www.workbuddy.cn/

WorkBuddy 文档站 https://www.workbuddy.cn/docs/workbuddy/From-

academic-research-skills https://github.com/Imbad0202/academic-research-skills

图片

图片

图片

图片

图片

图片

图片

图片

图片

图片

图片

图片

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry