maochiy/dsh-taskboard-plugin
@xiuda-ai/dsh-taskboard-plugin adds a task board to DeepSeek Harness: local JSON storage, a six-column board (Todo / In Progress / Blocked / In Review / Done / Archived, each column scrolling independently), drag-and-drop ordering (drop onto column whitespace or before a card via a floating sortOrder), task <-> session linking (a "Start conversation / View conversation" button that opens a DSH session), a task editor (title, description, status, priority, labels, assignee user/Agent, start/end dates; archive/restore/delete), multiple projects with a built-in global project (local) and an aggregate global view, and model-facing board tools (taskboard_list / taskboard_create / taskboard_update / taskboard_archive) so an Agent can manage the board directly. Logic is distilled from Proma's task board (floating sortOrder, optimistic locking, identifier numbering, localized status/priority tones).
Install
npm install @xiuda-ai/dsh-taskboard-pluginRun npm install @xiuda-ai/dsh-taskboard-plugin inside the target profile directory (~/.dsh/profiles/web), then append the taskboard insert row (name '@xiuda-ai/dsh-taskboard-plugin') to the profile's cordis.patch.yml, restart the harness to load the host half, and refresh the page for the browser half. The npm package @xiuda-ai/dsh-taskboard-plugin 0.1.0 was re-verified live on the registry on 2026-09-30 with its repository field back-linking to github.com/maochiy/dsh-taskboard-plugin (MIT). A local-path (symlink) option and a build-from-source flow are also documented. Data lives in ~/.dsh/taskboard/{projects.json, tasks.json} and is kept on uninstall by default.
Compatibility
DeepSeek Harness (the README assumes the Electron desktop app) with a target profile (default web). Runtime dependencies @deepseek-ai/cordis, @deepseek-ai/dsh-tools, and @deepseek-ai/dsh-home-paths must resolve from this package; for local development, link them to the profile's installed copies to match the running harness version. Storage is local JSON under ~/.dsh/taskboard/ with atomic writes (tmp + rename) and optimistic locking (version); no database. The session-header entry button only appears once you are inside a session.
Details
- Repo: maochiy/dsh-taskboard-plugin
- Category: Files & Data
- Stars: 1
- Version: npm @xiuda-ai/dsh-taskboard-plugin 0.1.0 (registry re-verified 2026-09-30); MIT
- Last push: 2026-08-14
- First seen: 2026-08-14
Recent updates
The README documents the feature list, the npm and local-path install options plus a build-from-source flow, the post-install usage (a Task Board pill in the session header; + 新建任务 creates a task), uninstall steps, the directory structure, the data storage layout, the internal HTTP API (POST /taskboard/api with ops state / listProjects / createProject / listTasks / createTask / updateTask / moveTask / archiveTask / restoreTask / deleteTask, with error codes such as VERSION_CONFLICT, PROJECT_NOT_FOUND, TASK_NOT_FOUND, INVALID_*), a Proma correspondence table, and limitations (v1 drops comments/timeline/relations/attachments/recurring tasks/undo; the backlog and canceled statuses exist in the data layer but are not rendered).
FAQ
- How do I install the DSH task board plugin?
- Per the README's Option 1: cd ~/.dsh/profiles/web then npm install @xiuda-ai/dsh-taskboard-plugin, append the taskboard insert row to cordis.patch.yml, restart the harness, and refresh the page. The npm package was re-verified on 2026-09-30 with its repository back-linking to maochiy/dsh-taskboard-plugin.
- Where is the board data stored?
- Locally in ~/.dsh/taskboard/projects.json and tasks.json, per the README, with atomic writes (tmp + rename) and optimistic locking; there is no database and the data directory is kept by default on uninstall.
- Can the agent manage tasks too?
- Yes -- the README lists model-facing tools taskboard_list, taskboard_create, taskboard_update, and taskboard_archive, so an Agent can read and manage the board directly.
Alternatives
cloader/dsh-taskboard · DamonKoy/dsh-web-ui#dsh-task-board · february2015/dsh-taskswarm