Introduction

In the DeepSeek Harness (DSH) ecosystem, plugins serve the role of extending model capabilities. For AI agents that need to perform multi-step tasks or handle complex workflows, maintaining contextual memory is critical. The xby-todolist plugin provides external working memory and task management features for large language models and AI agents, helping solve the problem of agents easily losing track of context or steps during long conversations or complex tasks.

Plugin Overview

xby-todolist is a DSH plugin maintained by xby-skill and categorized under “Memory.” It provides agents with structured task management services, helping them track progress, manage priorities, and maintain working memory. This supports reliable execution of complex multi-step tasks.

Core Features

The plugin provides the following core capabilities:

  • set_xby_apikey: Sets the API key in chat. This operation is automatically persisted and remains valid after restarting the process.
  • todo_read: Reads the current to-do list and returns the status, priority, and content of tasks.
  • todo_write: Creates and manages structured task lists, supporting real-time tracking of task status.

Installation and Activation

Install the plugin via the command line:

dsh plugin --profile web add github:xby_skill/xby-todolist

After installation is complete, restart the dsh process to activate it.

Typical Usage

Depending on the use case, invoke the tools in the following ways:

  1. At the beginning of a conversation, use todo_read to review pending items and confirm the current progress.
  2. Before starting a new task, use todo_read to prioritize existing work.
  3. When a user provides multiple tasks, use todo_write to create a list.
  4. When completing a complex multi-step task, use todo_write to track progress.
  5. Before starting work, mark the task status as in_progress.

Applicable Scenarios and Notes

  • Applicable scenarios: Primarily suitable for scenarios that require executing complex multi-step tasks or when a user provides multiple task instructions. The plugin can help agents maintain contextual consistency.
  • Usage notes:
    • todo_read does not accept any parameters; the input must be empty.
    • todo_write is not recommended for a single simple task.
    • Only one task can be in the in_progress status at a time.
    • Only mark a task as completed after it has been fully completed.

Summary

By introducing an external memory layer, xby-todolist improves the reliability and confidence of AI agents when handling complex tasks. The plugin is available under the MIT license and can be integrated into projects with confidence. For more information, visit the GitHub repository or the community catalog.