TutorialsAgent Automation: From Chatting to Getting Work Done
Three hands-on runs along one path — let an Agent finish a task on its own, wrap the method into a Skill, then schedule it to run every day.
This course does not teach you how to build an Agent. It teaches you how to hand over that one task you repeat every day and let the Agent run it to the end. Three hands-on runs form one path: get it done once, wrap the proven method into a Skill, then schedule it. One case runs through the whole course — a daily AI industry briefing.
What you take away
| Deliverable | What it is |
|---|---|
| 1 prompt formula | Role, task, context, format, constraints — five parts that make a complete job describable |
| 1 finished result | The Agent searches, filters, organizes and writes the file itself; you only verify |
| 1 Skill of your own | The proven method packaged as a reusable skill, so a new topic needs no new prompt |
| 1 scheduled task | Runs the Skill at a fixed time every day while your computer is on |
Problems this course solves
- Chat only: it answers, then you copy, paste and save the file yourself — nothing is saved;
- Rewriting the prompt every time: the same job done ten times means the prompt written ten times;
- No reuse: last time went well, but a new topic means teaching it from scratch;
- Knowing it should be automated but not where to start: you want it running daily, yet worry it might delete files or send something it shouldn't.
What you will be able to do
Describe a job in one formula
Fill in role, task, context, format and constraints, and the Agent no longer has to guess what to deliver, for whom, or where to save it.
Watch a real autonomous run
See it plan steps, call search and file tools, and actually land a file in the workspace instead of saying done in the chat box.
Keep the flow running
A Skill makes it reusable, a scheduled task starts it automatically — from then on you only verify.
Who this is for
Course structure
Lesson 1 · From chat to work: how an Agent differs from ordinary AI, how the four core concepts fit together, eight high-frequency scenarios, and the universal prompt formula.
Lesson 2 · First hands-on run: set up a workspace, hand over the prompt, let it search, organize and save by itself, then verify the result item by item.
Lesson 3 · Wrap and schedule: turn the working flow into a Skill, test it on a new topic, schedule it daily, and hold three quality red lines.
Materials
The full course handout is available to download or read online:
- Agent自动完成任务-课程讲义.md: full handout with all three hands-on prompts, five office scenario templates, a blank task design sheet and the quality red lines
- Handout online (Feishu): same content as the download, convenient on mobile
All hands-on work uses Tencent WorkBuddy. Installation and sign-in are covered in Lesson 2, step 0.
The one principle
You tell the Agent what to accomplish; the Agent works out how to accomplish it.
Treat it as a very capable assistant — you assign the work, supply the material, check the result and make the final call, rather than performing every step for it. The test is simple: would you use what it handed back as is? If not, a section of your task description is missing.
Three red lines before you start:
- It fabricates confidently — verify numbers, dates, names and sources yourself, especially in research tasks that touch the web;
- Confidential material stays out of the chat — desensitize client data, trade secrets and internal files first, and check which folders and systems it can reach;
- Output is not final — test-run a scheduled task, then check its results regularly; keep human confirmation for high-risk actions such as deleting files or sending messages.
Build Your Own First Small Tool
Choose your own topic, fill the three lists, build it with Baidu Dumate, fix at least one problem, and use it for one real task.
From Chat to Work: What an Agent Really Is
Ordinary AI answers questions; an Agent finishes the job. Learn the four core concepts, eight high-frequency scenarios, and the five-part prompt formula.