How to use dsh-superagent-prototype: setup, checks and limits

What dsh-superagent does

The agent is built for one GitHub repository. When someone allowed asks it for work, it opens a work item, works on it in its own agent session and directory, and answers in the original conversation, for example with a pull request.

A web panel shows what needs a person's attention. The project's stated principle is to take as little human attention as possible while giving as much value as possible, so the agent acts only when asked or when a rule the team wrote allows it.

• Listens to GitHub and, optionally, Feishu
• Opens a work item for each request from an allowed person
• Works in its own agent session and directory
• Replies in the original conversation, for example with a pull request
• Shows items needing a person in a web panel

Get the code and install it

The README lists clone-and-build as the install method and does not list prerequisites. Clone the repository from GitHub with the command below.

Setup has many steps, so the README recommends letting a coding agent guide you. Ask your agent to read SETUP.md in the repository and set up dsh-superagent with you. The agent asks one short question at a time, shows the full plan once, then installs and configures everything. For a manual route, the README points to docs/006-deployment.md, which covers manual setup, configuration, upgrades and troubleshooting.

• Clone the repository
• Read SETUP.md, or ask a coding agent to follow it with you
• For manual setup, read docs/006-deployment.md

Run the project checks

Once the code is in place, the README's development command runs the typecheck, the tests and the build in one step. Use it to confirm the project works in your environment before you adapt it.

The README asks you to review and adapt the code before using it in your own environment, because it may contain bugs and its practices may be inconsistent.

Read the documentation in order

The numbered documents build on each other, so read them in sequence. Start with the philosophy and domain model, then authorization and execution, before moving to deployment and architecture.

The list covers: 000 Philosophy, 001 Domain model, 002 Authorization, 003 Execution, 004 Prompts, 005 Panel, 006 Deployment, 007 Architecture and 008 Cookbook, which covers adding a Source.

• 000 Philosophy: the principle and what it asks of the operator
• 001 Domain model and 002 Authorization: how work items and permissions fit together
• 003 Execution and 004 Prompts: how the agent works and what it is told
• 005 Panel: the web review panel
• 006 Deployment: manual setup, configuration, upgrades and troubleshooting
• 007 Architecture and 008 Cookbook: system design and adding a Source

Know the limits before you adopt it

The README describes this as a research prototype, not an official DeepSeek product, and not production-ready. Some designs in the accompanying article are future directions and are not implemented in this prototype, and the code may not match the internal version the team uses.

It is a good fit for studying how work items, authorization and a review panel can work together. It is a poor fit if you need a quick install or dependable production software.

• Not an official DeepSeek product; research prototype
• Some article designs are not implemented
• Setup has many steps; not a quick install
• The code may differ from the internal version