Templates and daily work

Based on the 3 November 2026 integrated edition · iPolloWork 0.50.33 · Download the complete Chinese PDF or Word manual

Screenshots show the source application in Simplified Chinese. New task, schedule and editor examples use sample data; verify actual saves, connections and outputs in your project.

Use, import and save templates

Open Templates. Packages use .ipwp, with legacy .ipwt support.

1. Preview a category, install the template and then use it. Import local packages from the import action.

2. Edit the independent result and preserve reusable variables and assets before saving it as a personal template.

3. For enterprise sharing, select the submission destination, add a version and wait for review. Update or uninstall from the template card.

Completion check: The template is listed under the right source and a new result does not overwrite the original.

Plugins, Skills and MCP

Check the installed capability before adding another one.

1. Install and enable the required extension, complete its configuration and authorize external accounts when prompted.

2. Add MCP through the current engine configuration: name, URL, OAuth and project or global scope.

3. Use / to find a command or Skill and @ to reference a capability. Run one small task that exercises the tool.

Completion check: The capability is discoverable and an actual call succeeds. Use plugin tutorials for individual services.

Create, import, export and update standalone Skills

1. Open Skills and inspect its personal, public or enterprise source before installing.
2. Create a Skill with its purpose, steps, inputs and completion checks, or import a supported file. A standalone .md import is limited to 1 MB; .ipollowork-skill packages can carry related resources.
3. Review required resources and permissions, then test the Skill on a small real task.
4. Export the appropriate format for sharing. Review source and version before updating, and revoke unnecessary access or uninstall when no longer needed.

Completion check: A new task discovers the intended version and can follow its resources. Imported instructions do not grant permissions to external services.

Browser and media work

Use the right-side Web panel and Settings → Authorization center.

1. On desktop, open a Web tab and URL. For automation, specify the target, account scope and allowed actions.

2. Configure a supported media service before asking for transcription, translation, speech or video generation.

3. Inspect asynchronous job status and outputs, then save the generated files. Preserve temporary media URLs as required by the task.

Completion check: A visible browser effect or real media output exists. The built-in browser is desktop-only; continuous simultaneous interpreting is outside this complete workflow.

Computer use and browser connections

1. Choose browser automation for webpage tasks and the configured computer-use tools for supported desktop actions.
2. Check the required MCP connection, installed capability and operating-system permissions. Browser account sign-in is a separate requirement.
3. State the target window or website, intended account and allowed actions. Inspect a small real action before expanding the task.

Completion check: The operation affects the correct surface and account. A connected browser does not automatically establish desktop control or system permission.

Save outputs to object storage

Configure OSS or Wasabi and the intended bucket in Authorization center.

1. Set the provider and default routing, then identify the workspace files to upload.

2. Explicitly request uploading the selected files and returning an access method. Ordinary outputs can remain local.

3. Check the object and link permissions and open the local deliverable from the file or output panel.

Completion check: Both the local file and uploaded object can be verified. Keep storage keys out of templates and deliveries.

Search, export and schedule entry points

Use Ctrl/Cmd + F within a conversation and Ctrl/Cmd + Shift + F across conversations; export Markdown from the task menu. See conversation history for details.

Open Schedule to plan dated work. Ordinary calendar entries record a plan; enable and verify automatic execution separately. See the following sections for schedule creation and natural-language previews.

Command palette, shortcuts and help

1. Open the command palette with Ctrl/Cmd + K and choose the intended action.
2. Use Ctrl/Cmd + N for a new task when focus is outside an editable field, Ctrl/Cmd + J for the local terminal, Ctrl/Cmd + F for conversation search and Ctrl/Cmd + Shift + F for global search. Check platform-specific labels.
3. Open Help, search or choose a category, then follow the anchored answer for the current feature.

Completion check: The command operates in the intended scope and the help result matches your installed version.

Schedules and planning

Entry: Schedule in the sidebar, or the schedule view inside project tasks.

1. Choose the date and available day, week or month view. Check whether the scope is global or project-specific.
2. Create an entry from the available action or time slot and select the target project.
3. Set the title, description, start, due time, status, priority and owner as needed. The due time must not precede the start.
4. Save and reopen the entry. When cancelling, check whether automatic execution also needs to be disabled.

Completion check: The date, project and task detail agree. A calendar item, due time or status change alone does not run an AI task.

Scheduled and recurring automatic execution

Entry: Timing and automatic-execution controls in a schedule or task detail.

1. First run a similar task manually, with the intended project, engine, tools, files and completion conditions.
2. Set a valid start time and enable automatic execution. Choose once, daily or weekly, and follow the project agent’s model or select a connected model.
3. Save and inspect the plan. Keep the desktop application or project server running, with network access, credentials, quota and file paths available.
4. At the planned time, inspect last-start information, the linked conversation and actual outputs. Resolve failures before testing again.
5. To pause or cancel, disable automatic execution and separately stop any conversation that has already started when necessary.

Completion check: The linked conversation contains a real run and output, and the next recurrence is correct. The running service handles timing; second-level precision, execution while shut down and every missed run during sleep are not guaranteed. A due time does not force a run to stop.

Create schedules through conversation

1. In a task with schedule-tool permission, specify the target project, activity, date, time, time zone and duration. Include daily or weekly recurrence when needed.
2. State whether this is a plan only or should automatically execute, including required input and output files.
3. Review the returned preview for time zone, start, end, recurrence and automatic-execution state.
4. Confirm the preview before writing the entry, then reopen it in Schedule.

Completion check: The confirmed entry matches the preview and intended project. An ordinary calendar plan and an automatically executed task are clearly distinguished.

Remote workspaces and connection checks

1. Prepare the remote workspace service address, required credentials and accessible project directory.
2. Connect through the available workspace entry, check the target and test a small task with actual file access.
3. After reconnecting, inspect the session and output state before retrying actions that may already have completed.

Completion check: The intended remote project executes the task and persists its outputs. A remote execution service and a model inference service have different responsibilities. Desktop-only Design, browser and terminal paths require separate availability checks.

Next: continue reading, or browse the complete feature index.