Meowbert Docs

Appearance

Tool Options Reference ​

The Task Composer's wrench menu lets you enable optional capabilities for a task. This page describes each option in detail.

Memory ​

What it does: Allows the agent to semantically search the workspace's persistent .memory directory during the task. Search results include snippets plus file paths and line ranges so the agent can inspect or update the underlying note.

When it appears: Only in workspaces where Memory is enabled in Workspace Settings.

When to enable it:

  • The task should reuse durable project notes, decisions, or preferences.
  • You want the agent to find the right Memory note before editing it.
  • The workspace keeps long-lived references in .memory.

When to leave it off:

  • The task is fully self-contained and doesn't need prior durable notes.
  • You want the run to ignore existing Memory and work only from the prompt plus attached files.

See Workspace Memory for how .memory is organized.

What it does: Allows the agent to fetch real-time information from the internet during the task. The agent can search for current documentation, news, pricing, API references, or any other web content.

When to enable it:

  • The task requires information that might have changed since the model's training cutoff.
  • You want the agent to verify or cross-check facts against live sources.
  • The task involves current pricing, availability, or status information.

When to leave it off:

  • The task is self-contained (e.g., transforming files you've provided).
  • You want fully reproducible, offline results.
  • Latency is a concern — web search adds round-trip time.

Schedule Tasks ​

What it does: Converts the task into a recurring scheduled task that runs on a defined interval. When enabled, a schedule picker appears in the composer where you can configure the frequency (e.g., every hour, daily at 9am, every Monday).

When to enable it:

  • You want automated, periodic execution of a prompt (e.g., a daily report, a nightly cleanup job).
  • You need the agent to check something on a regular schedule and notify you of changes.

Limits: Each project has a maximum number of active recurring tasks (configured by platform admins). Check your workspace limits if you hit an error creating a scheduled task.

Managing scheduled tasks: Scheduled tasks appear in the Project Overview with the scheduled type badge. Cancel them from the overview to stop future runs.

Subtasks ​

What it does: Allows the agent to create child tasks for scoped portions of the work. The parent task delegates to subtasks, waits for them to complete, and aggregates results.

When to enable it:

  • The work is naturally parallelizable — for example, processing multiple files independently.
  • The task is so large that breaking it into smaller pieces makes each more reliable.
  • You want cleaner separation of concerns in the event log.

When to leave it off:

  • Simple, single-step tasks — the overhead isn't worth it.
  • You want a single coherent event stream without branching.

Skills ​

What it does: Enables one or more Skills for the task. Skills are MCP (Model Context Protocol) tool packs that extend the agent with specialized capabilities beyond shell commands. When a skill is enabled, its tools are available to the agent during the task.

See the Skills Reference page for descriptions of all available skills.

When to enable it:

  • The task requires browser automation (use the Browser Use skill).
  • The task needs to generate or manipulate Word documents (use DOCX Studio or OpenAI: doc).
  • The task needs to generate or manipulate PowerPoint files (use PPTX Studio or OpenAI: slides).
  • The task needs spreadsheet analysis or workbook editing (use OpenAI: spreadsheet).
  • The task needs PDF generation, extraction, or visual review (use OpenAI: pdf).

Note: Skills must be installed and configured on the server. Available skills vary by platform deployment. Contact your platform admin if a skill you need isn't listed. You can enable more than one skill for the same task when the workflow spans multiple artifact types.