Scheduled Tasks: Let WorkBuddy Work on Its Own, On Time
The previous lesson's experts solved "the quality of one-off tasks"; this lesson solves "the manpower of repetitive tasks". WorkBuddy Scheduled tasks are an official automation capability: task name, prompt, execution time, workspace, permissions, and other configurations are packaged and saved, and when the execution time arrives, the task is automatically initiated with your current logged-in identity—daily briefings, weekly report summaries, scheduled data cleanup, none of which you need to trigger manually anymore. Item 9 of the official 10 getting-started tips is about exactly this: "Make good use of automation to let WorkBuddy keep working on its own" and "Let WorkBuddy run around the clock, 24 hours a day". In 12 minutes, set up your first task that runs on its own on schedule.
This is lesson 3 of "Efficient Workflows", based on WorkBuddy v5.9.3, with content drawn from the official "Scheduled Tasks" feature documentation.
- Just jumped into this section? Start with the roadmap in 3.0 Efficient Workflows: From Basic Use to Great Use
- For how to write prompts clearly, see the three-element formula in 3.1 The Requirements-Setting Methodology—scheduled tasks only help you "send on time", not "write well"
What You'll Be Able to Do After This Lesson
- Create a WorkBuddy scheduled task using three methods: manual creation, creating from a task, and creating from a template
- Configure the 7 recurrence rules so the task runs automatically on schedule, e.g., every Friday at 17:00 or every morning at 8:30
- Explain the differences between the three task sources—local, project, and cloud—cloud tasks run on time even when your computer is off
- Choose the right execution permission for unattended tasks so they don't get stuck "waiting for confirmation"
- Know how to view run history, track down failure reasons, and push task results to WeCom
Your Current Struggles
- Every morning you manually have AI search the news and put together a morning report; miss a day and the streak breaks
- At the busiest moment before leaving work on Friday, you still have to rack your brain to summarize the week's work and write the weekly report
- You want AI to work on a schedule, but worry it might mess with files when running automatically with no one watching
When to Use This Technique
When you need to:
- Hand highly repetitive, clearly rule-based work that needs no real-time human intervention over to WorkBuddy for automatic handling—this is the official "suitable for automation" criterion
- Receive a morning report at a fixed time every day and a weekly report at a fixed time every week
- Have tasks run on schedule even when your computer is offline (using cloud tasks)
And you don't want to:
- Manually copy and paste the same prompt every day
- Personally sit at the computer at 17:00 on Friday to aggregate data
- Crawl out of bed in the middle of the night to click "Execute"
🎒 Preparation Before Starting
- Have read 3.0 Efficient Workflows: From Basic Use to Great Use and know the roadmap of this section's 4 lessons
- Can write prompts using the three-element formula from 3.1 The Requirements-Setting Methodology—scheduled task prompts follow the same approach
- The WorkBuddy client is logged in and your credit balance is positive—every scheduled task run consumes credits, see 2.6 Credits and Pricing
- Know your automation task quota: 3 for the Trial plan, 15 for Standard, 30 for Advanced, and 99 for Flagship (the first three tiers are officially marked as "99 free for a limited time"; the official pricing page prevails)
- Back up first when tasks involve real files—the official recommended order, verbatim: "Back up before trying remote and automation when real files are involved: steady first, then fast"
- 12 minutes of reading time; ideally, create a task and follow along as you read
Core Idea
Scheduled tasks solve the problem of "having to trigger things manually every time".
The official definition in one sentence: save the task name, prompt, execution time, workspace, permissions, and other configurations, and when the execution time arrives, the task is automatically initiated with your current logged-in identity to complete queries, summaries, file processing, and other work per the prompt. You can understand it as: package up "prompt + schedule + permissions" and run it automatically on time. Tasks only execute when triggered by the configured time rule, and are subject to task frequency, maximum execution duration, and system concurrency limits.
Scheduled tasks come in three sources—understand the differences before you start:
| Source | Where It Runs | What It's For | Watch Out For |
|---|---|---|---|
| Local | Local client | Tasks that need to read/write local files | Do not shut down your computer or exit the client during execution |
| Project | Inside the corresponding project | Periodic work within a project | View-only on the scheduled tasks page; edit and pause must be done in the corresponding project |
| Cloud | Cloud sandbox | Tasks that must run even when the computer is off/offline | Cannot access local files, no WeCom push support; cards carry a cloud badge |
What work is suitable for automation? The official criterion: highly repetitive, clearly rule-based, no real-time human intervention needed. Official examples: automatically generate a daily work report at 9:00 every day, automatically run tests after code commits, aggregate data and generate a report every week, periodically check code standards, and periodically fetch and organize industry news.
Follow Along
Step 1: Open the scheduled tasks page and get to know the three task sources
Why: Know the entrance before you create anything. Scheduled tasks have their own page, where the three sources—local, project, and cloud—are managed together. Don't discover you "created it in the wrong place" after building a task.
Steps:
- Open the WorkBuddy client and click "Scheduled Tasks" in the left navigation bar.
- Get to know the page's two tabs: Scheduled Tasks (view and manage all tasks) and Run History (the past runs of all tasks).
- On your first visit, the page will recommend common task templates—give them a quick look; you'll use them in Step 2.
You should see: each task card shows its execution rule and next run time; for tasks whose last run failed, a red failure badge appears next to the title—hover over it to see the reason.
- I found the scheduled tasks entrance and can explain what the "Scheduled Tasks" and "Run History" tabs each manage
Step 2: Create your first WorkBuddy scheduled task
Why: There are three official creation methods, matching three situations—when you need precise control, when you want to save effort, and when you don't know what's possible. Click "New" in the top-right corner of the page to create one manually; click the dropdown arrow next to it to see all three methods.
Method A (recommended): Create manually
Best for: when you need to precisely specify the execution time, permissions, and push channel.
- Click "New" and fill in the configuration in the editing window that pops up (the meaning of each field is covered in Step 3).
- Note: You can only click OK after filling in both the name and the prompt.
Method B: Create from a task
Best for: describing your need in one sentence and letting WorkBuddy create it for you.
- Click the dropdown arrow next to "New" and choose "Create from Task".
- WorkBuddy will return to the home page with "Help me create a scheduled task:" pre-filled in the input box—add your specific requirements after it and send. Official example:
Help me create a scheduled task: every weekday at 9 a.m., compile yesterday's sales data and send it to the product group chat- The creation result will be shown in the conversation, and you can click to jump to the task details.
Method C: Create from a template
Best for: when you're not sure what's possible and want a quick start from common scenarios.
Choose "Create from Template", or click a recommended task template on the blank page (such as daily AI news push, weekly work report, or pre-meeting prep)—the editing window will be pre-filled with the name, prompt, and execution time; modify as needed and click OK.
You should see: with manual creation and the template method, the task appears in the scheduled task list after you click OK; with the create-from-task method, the result is shown in the conversation, and you can click to jump to the task details.
💡 You can create one without visiting the scheduled tasks page
In a regular conversation, just describe a recurring need (e.g., "Remind me every morning at 8 a.m. ...") and WorkBuddy will likewise create a scheduled task for you.
- I created a scheduled task using one of the three methods (for this walkthrough, Method A or Method C is recommended)
Step 3: Fill in the Prompt, Recurrence Rules, and Workspace
Why: Configuration items determine what a task does, when it runs, and where files go. In the manually created editing window, fill in the following:
| Configuration Item | Description |
|---|---|
| Name | Used to distinguish different scheduled tasks |
| Prompt | Describes the task's goal and output requirements |
| Model, experts, skills, connectors | Selected at the bottom of the prompt input box; click + to add experts, skills, and connectors |
| Workspace | Specifies the task's execution directory and where files are saved |
| Permissions | Default permissions or full access permissions (detailed in Step 4) |
| Repeat | Execution frequency, with interval, one-time, daily, weekly, biweekly, monthly, and yearly as the 7 options |
| Time | The specific time of each run |
| Valid period | The date range during which the task is active (not needed for one-time tasks) |
| Push messages to WeCom | Pushes the result to WeCom after the task completes (detailed in Step 6) |
Three details about repeat rules:
- For weekly / biweekly, you can select multiple days of the week at once, e.g., every Monday, Wednesday, and Friday
- For monthly, you can select multiple dates at once, e.g., the 1st and 15th of each month
- Configure multiple execution dates for the same task in a single rule—do not split them into multiple scheduled tasks
Not sure how to write the prompt? Just copy the two automation scenarios from the official 10 tips:
Scenario 1: Daily morning briefing
Set up a scheduled task that runs every morning at 8:30: send a daily news briefing to my email inbox.
Scenario 2: Weekly report auto-generation
Run automatically every Friday at 17:00: compile all code commit records under D:/项目/ from this week,
group them by module, count the number of commits and main changes per module, and generate a concise weekly report.The briefing goes to email, so authorize the QQ Mail connector first; the weekly report reads a local directory, so remember to choose the right workspace. For full walkthroughs of both scenarios, see A1 Weekly Report Auto-Generation and C4 Daily News Briefing.
💡 The workspace also affects where the entry appears
Local scheduled tasks take up only one entry in the left navigation bar (with a scheduled tasks icon): with no workspace selected, they appear in the task list; with a workspace selected, they appear under that workspace.
You should see: after saving, the task card shows the execution rule and next run time you configured—check the card to confirm the date and time are correct.
- My task has its repeat rule and time configured, and the prompt is written clearly using the three-element formula
Step 4: Set execution permissions—choose "Full Access Permissions" for unattended runs
Why: Scheduled tasks run automatically when the time comes, and at that moment you're most likely not at your computer. With the wrong permission setting, the task will "stop halfway."
| Permissions | Description |
|---|---|
| Full access permissions (recommended) | The task runs unattended and tool calls don't require confirmation each time; enabling this requires an extra confirmation |
| Default permissions | Sensitive operations require your confirmation; if you're not at your computer, the task will halt in a waiting-for-confirmation state |
The two permission tiers work the same as for regular tasks; for background, see 2.4 Three Execution Modes.
You should see: when you choose full access permissions, the system will ask you to confirm once more—this is an intentional safeguard against accidental activation; only after confirming can the task run smoothly unattended.
⚠️ Two safety boundaries
- Local scheduled tasks run in the local client; do not shut down your computer or exit the client while a task is running, or it may be interrupted.
- Tasks run automatically unattended, so set prompts involving file writing / deletion / financial operations with care (quoting the official safety reminder).
- My unattended task uses full access permissions, and its prompt contains no dangerous file deletion or financial operations
Step 5: Do a trial run first and check the results and failure reasons
Why: The first item of the official usage advice—after creating a task, click run to do a test run first, confirm the execution result meets expectations, and only then scale up the execution frequency. Going straight to high frequency every day means burning credits for nothing if it goes off track.
Steps:
- Click the Run button on the task card to execute it once immediately, without affecting the original schedule.
- Click the local task card (or the scheduled task entry in the sidebar) to enter that task's run page and view the results.
You should see: past runs displayed on a timeline, with the latest at the bottom; the page opens the artifacts of the latest run by default; the top shows the next run time; clicking a run takes you into the corresponding session, where you can view the full execution process and continue the conversation.
Where to check when a run goes wrong? The Run Records tab:
- Each record contains the task name, status, run time, duration, and task type, and can be filtered by status (success / failure / running), task type (local / project / cloud), and specific task
- When a model queuing failure occurs (e.g., the current model's queue is full), the task does not fail immediately; it is automatically re-queued and retried; it is only marked as failed after multiple retries still fail
- When a run fails, click the record to view the failure reason and handling suggestions; reasons officially listed: connector not authorized, login expired, computer offline causing the execution time to be missed, model queue retries exhausted
- Records no longer needed can be archived or deleted; archived records can be viewed in Settings > Data Management
Other common actions on the task card:
| Action | How | Result |
|---|---|---|
| Edit | The edit button on the card | Modify the task configuration |
| Pause / Enable | The toggle on the right side of the card | Pausing only affects subsequent executions; the currently running one will not be interrupted |
| View run records | More (⋯) → View run records | See only the past runs of this task |
| Open task | More (⋯) → Open task | Enter that task's run page |
| Delete | More (⋯) → Delete | After deletion the task no longer executes, this cannot be undone |
| Batch management | Batch management in the top-right corner | Check multiple tasks and delete them together |
Filtering and search: filter by status (all, enabled, paused); click "Filter" to filter by source (local / project / cloud); click the search icon to search by task name or run records.
💡 Results will proactively notify you
- Desktop: after enabling Settings > General > Scheduled Task Artifact Notifications, you will get a system notification when a task generates artifacts; clicking the notification opens the corresponding artifact directly.
- Mobile: once the WorkBuddy App is linked to the current computer, it receives notifications when a scheduled task completes, awaits confirmation, or fails; the title starts with "Scheduled task:".
- I did a test run of a task and can understand each run's status and failure reasons in the run records
Step 6: Push results to WeCom
Why: You're not at your computer when the task finishes—push lets you receive the results directly in WeCom. Pushing uses the WeCom bot already bound to the assistant; for background on the assistant and its channels, see 5.1 Assistant Overview.
Steps:
- In the bottom-left corner of the create or edit window, turn on the "Push messages to WeCom" toggle.
- The assistant is already bound to a WeCom bot via a WebSocket persistent connection → the toggle turns on directly, no other action needed.
- Not yet bound → next to the toggle it says "No WeCom bot connected yet; please complete the connection first"; click "Connect", then click "Get WeCom QR code" in the pop-up window.
- Scan the QR code with WeCom to complete the bot onboarding; if you don't have a WeCom account, click "Register" below the QR code.
- After a successful binding it says "Binding successful, one step to go": open WeCom, send any message to this bot to complete the final confirmation, then click "Sent".
You should see: back on the scheduled tasks page, the push toggle is on automatically and the task content you filled in is not lost; when the task completes, you receive the results in WeCom.
⚠️ Two pitfalls with push
- If you have never sent a message to the bot in WeCom, the scheduled task cannot find a target to receive the push, so results cannot be pushed—after binding, be sure to complete this final "send a message" step.
- Scheduled task push** only supports WebSocket persistent connections, not URL callbacks**. If the assistant is currently connected via URL callback, turning on the toggle prompts "URL callback connections do not support message push yet; you need to unbind and reconnect"; reconnecting cancels the existing URL callback, and the assistant's original ability to receive messages via URL callback is cut off accordingly.
Two more official notes: for scheduled tasks that previously pushed via a "standalone WeCom bot", the original push method remains unchanged; push failures do not affect the task's own execution.
- I know push only supports WebSocket persistent connections, and I completed the final "send a message to the bot" confirmation
For details, refer to the official documentation: WorkBuddy Scheduled Task Feature Description.
Checkpoint ✅
- Can explain the differences among the three task sources: local executes on your machine, project requires managing in the corresponding project, and cloud executes in a cloud sandbox and can run even when your computer is off
- Created a WorkBuddy scheduled task using one of the three methods, and knows the templates include daily AI news push, weekly work report, and pre-meeting preparation
- Can configure multi-date rules like "every Monday, Wednesday, Friday" or "on day 1 and day 15 of each month", and knows to finish configuring within one rule instead of splitting into multiple tasks
- Enabled full access for unattended tasks, and knows not to shut down the computer or exit the client while a local task is executing
- After creating a task, ran it once immediately as a test run, and can read the failure reasons in the run records
Pitfall Reminders
❌ Computer shut down, local task didn't run
Symptom: On Friday at 17:00 the person was off work and the computer was shut down; the weekly report task did not execute.
Cause: Local scheduled tasks execute in the local client; do not shut down the computer or exit the client during execution; "computer offline causing the execution time to be missed" is exactly one of the officially listed failure reasons.
Correct approach: Use cloud tasks for work that must "run even when shut down"; for local tasks, make sure the computer and client are online during the execution window.
❌ Ran an unattended task with default permissions, stuck at "awaiting confirmation"
Symptom: The task started on schedule but stayed in the awaiting-confirmation state and never proceeded.
Cause: Under default permissions, sensitive operations require your confirmation, and if you're not at the computer, no one can confirm.
Correct approach: Choose full access for unattended scheduled tasks (officially recommended); at the same time, set prompts involving file writes / deletions / financial operations with care.
❌ Bound the WeCom bot, but the push still won't go out
Symptom: The toggle is on and the bot is bound, but no results arrive in WeCom after the task completes.
Cause: After binding, you didn't send any message to the bot in WeCom to complete the final confirmation, so the scheduled task cannot find a target to receive the push; or the assistant uses a URL callback connection, while scheduled task push only supports WebSocket persistent connections.
Correct approach: Follow the guide to send a message to the bot and then click "Sent"; for URL callback, unbind and reconnect as prompted.
❌ Wrote local file paths in a cloud task's prompt
Symptom: After the cloud task ran, it reported that files could not be found.
Cause: The cloud sandbox cannot access local files; the official docs explicitly say "do not reference local file paths in prompts".
Correct approach: Use local tasks with a properly chosen workspace for work that needs to read or write local files; schedule only work that doesn't depend on local files as cloud tasks.
Advanced Tips (Optional)
🔧 Manage scheduled tasks with the mobile App. The WorkBuddy App's "Automation" page can also create tasks: tap the + icon to create via a form (title, prompt, execution frequency, execution time), or tap "Create via chat" to describe in natural language, e.g., "Every morning at 9 o'clock, push a summary of yesterday's AI industry news". The ... menu in the top-right corner of the task list supports editing, test run (execute once immediately to verify), pause/resume, and delete; tap a task card to view historical execution records and test results. Note: this part is a guide for the App's "Cloud Work" mode; the automation under "Connect Computer" is for receiving execution results of computer-side tasks.
🔧 Four details about cloud tasks. ① They can still run on schedule even when your computer is shut down or offline; ② When editing, you don't need to select a workspace or permissions, and pushing to WeCom is not supported; ③ After deleting a cloud task, the cloud sessions it has already produced are still retained, and you can continue to view the process and outputs; ④ When offline, cloud tasks can only be viewed, and can only be operated after reconnecting to the internet.
🔧 Estimate credit consumption in advance. Every run of a scheduled task consumes credits, depending on task complexity: simple reminder tasks cost less, while code inspections, document summaries, connector analyses, and the like may consume more tokens. When a task calls a connector, the connector's own reads and writes do not consume credits; WorkBuddy credits are only consumed when understanding and summarizing external data. For plans and add-on packs, see 2.6 Credits and Pricing.
🔧 For weekly and morning reports, copy the official scenarios directly. The official 10 tips provide two ready-made scenarios: every day at 8:30 send a morning report by email, and every Friday at 17:00 summarize code commits in D:/项目/ to generate a weekly report. For full walkthroughs, see A1 Automated Weekly Report Generation and C4 Daily News Briefing.
Lesson Summary
✅ A WorkBuddy scheduled task = a bundle of prompt + schedule + permissions: save the configuration, and when the execution time arrives, the task is automatically initiated under your current logged-in identity.
✅ Three sources: local (executed on this machine, don't shut down), project (manage in the corresponding project), and cloud (cloud sandbox, runs even when shut down, but cannot access local files).
✅ Three ways to create: create manually (precise configuration), create from a task (one sentence), and create from a template (ready-made templates such as daily AI news push and weekly work report).
✅ 7 recurrence rules: interval, one-time, daily, weekly, biweekly, monthly, and yearly; configure multiple dates within a single rule.
✅ Choose full access permissions for unattended runs; after creating a task, do a trial run first, confirm the results, and then scale up the execution frequency.
✅ Results can be pushed to WeCom: only WebSocket long connections are supported; after binding, remember to send a message to the bot.
Review of key configuration fields:
| Configuration Item | Recommended Value | Description |
|---|---|---|
| Prompt | Write it following the 3.1 three-element formula | Describes the task goal and output requirements |
| Recurrence + Time | Choose one of the 7 rules as needed | For weekly/monthly multiple dates, configure them within a single rule |
| Workspace | For local tasks, select the execution directory | Cloud tasks don't need one and cannot read local files |
| Permissions | Full access (for unattended runs) | Default permissions will pause waiting for confirmation |
| Validity Period | The date range during which the task is active | No need to set for one-time tasks |
| Message Push | Enable when needed | WebSocket long connection only; send a message to the bot first |
Next Lesson Preview
Scheduled tasks let WorkBuddy "work on time"; in the next lesson, we'll make it "remember your materials"—put frequently used files into the knowledge base, let it remember your preferences, and have them available whenever a task runs. In the next lesson, you'll learn about 3.4 Knowledge Base and Memory: Making WorkBuddy Remember Your Materials and Preferences.
You will learn:
- How to import and manage files in the knowledge base so WorkBuddy can reference them directly
- How memories are generated and managed so it remembers your preferences and habits
- How other knowledge base capabilities such as the dedicated mailbox work together with tasks