AI Agent Hub
Back to plugins
⚙️

dsh-tasks-manager

Workflow Updated 2026.08.26

Run the following command in DeepSeek Harness:

dsh plugin install navid-kianfar/dsh-tasks-manager

Paste the following prompt into your AI chat to install this plugin:

Run `dsh plugin install navid-kianfar/dsh-tasks-manager` in your terminal to install this plugin; the source code is available at https://github.com/navid-kianfar/dsh-tasks-manager .

About this plugin

When coding with DeepSeek Harness, tasks tend to live in a session-scoped todo list that vanishes the moment you close the terminal. dsh-tasks-manager persists them in a SQLite file (.dsh/tasks.db) inside the repository, so the board travels with the code, survives across sessions, and needs no external SaaS.

The board centres on a five-column kanban (backlog / todo / in progress / blocked / done) with drag-and-drop reordering, filtering by status, priority, labels, assignee, and full-text search, and assignees auto-detected from git log and ranked by contribution. Crucially, the model can call task_add, task_update, and related tools directly, so a single natural-language sentence creates or edits a card. Pair this with subagent dispatch and a card can be handed off to the agent to work in the background, with live status, on-demand output, and a stop control right on the card. Archiving keeps the full history intact; deletion is a separate confirmed action the model cannot perform unless you opt in.

It suits individuals or small teams who work daily in the Harness and want a project board that lives next to the code, that the model can read and write to, without maintaining yet another external tracker.

Use Cases

  • Persist project tasks across sessions without external SaaS
  • Create or update task cards with a single natural-language sentence
  • Dispatch a card to a background agent and track its live status

Best For

  • Individual developers working daily in the DeepSeek Harness
  • Small teams that want the task board version-controlled with the code
  • Engineers who need the model to read and write tasks without manual CLI steps