dsh-task-panel
Run the following command in DeepSeek Harness:
dsh plugin install lihang-lh/dsh-task-panel
Paste the following prompt into your AI chat to install this plugin:
Install by running dsh plugin install lihang-lh/dsh-task-panel inside DeepSeek Harness; the source repository is https://github.com/lihang-lh/dsh-task-panel. After installing, restart dsh web and refresh the page to see the Task Panel button at the bottom of the sidebar (above Settings).
About this plugin
In DeepSeek Harness, chat bubbles are the only interface for tracking work. Tasks scatter across sessions, and you end up scrolling through history just to figure out who is at which step and what is still unverified. dsh-task-panel inverts that: you publish a task and accept the result, while planning, clarification, development, and review are all delegated to a subagent pipeline.
At its core is a seven-state pipeline (todo, clarify, confirm, develop, paused, review, done). Each stage carries a dedicated prompt with JSON Schema output constraints, and state transitions are driven by automatic readers rather than manual polling. On publish, the plugin full-text-searches your existing sessions and auto-attaches relevant context so subagents can pick up where you left off. Development supports pause, resume, and re-edit; review enforces a zero-issues quality gate and auto-retries with the issue list (up to three rounds) before the task can be accepted. The panel surfaces every task with color-coded tabs, status dots, checklist progress bars, screenshot thumbnails, and a pass-fail verdict banner.
It is built for developers who drive DSH daily and want a structured workflow layer on top of chat without maintaining their own state machine. Especially useful when a task spans multiple sessions, needs a clarification round between a rough direction and concrete steps, and requires traceable checklists and screenshot evidence at acceptance time.
Use Cases
- Publish a task and let subagents handle planning, development, and review end-to-end
- Link historical sessions to new tasks so subagents inherit prior context
- Auto-revert to development with the issue list when review finds unresolved issues
Best For
- Developers who use DSH daily for AI-assisted programming
- Individuals or teams needing a structured task layer on top of chat
- Users who want acceptance-driven workflows with less manual state tracking
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.