dsh-superagent-prototype
dsh-superagent-prototype: a proactive AI coding agent for a single GitHub repository, delivered as a DeepSeek Harness (DSH) bundle.
Install
dsh plugin --profile web add github:deepseek-ai/dsh-superagent-prototype What it solves
- Day-to-day collaboration on one repository (issues, pull requests, optionally Feishu messages) normally needs a human watching it; an agent cannot pick up work and drive it on its own.
- This prototype provides a listenable entry point: when an allowed person asks for work, it opens a work item, executes it in its own agent session and its own directory, and answers back in the original conversation — for example with a pull request.
- A web panel shows what needs a person, so an agent that can act for a whole team does not end up consuming the whole team’s attention.
- The stated principle: take as little human attention as possible and give as much value as possible. It acts only when asked or when a rule the team wrote allows it, and speaks only when someone must act or can use a result.
Who it’s for
- Teams and repository maintainers who want a proactive coding agent attached to their GitHub repo.
- Developers and researchers studying autonomous-agent design trade-offs: authorization boundaries, execution isolation, prompt organization.
- People already using DeepSeek Harness who want to see how a bundle and its authorization model look in practice.
Docs and setup
Read the numbered documents in order: 000 Philosophy, 001 Domain model, 002 Authorization, 003 Execution, 004 Prompts, 005 Panel, 006 Deployment (manual setup, configuration, upgrades, troubleshooting), 007 Architecture, 008 Cookbook (adding a Source). Setup is intended to be done with a coding agent following SETUP.md: one short question at a time, the full plan shown once, then install and configure.
Caveats
The authors state this is a research prototype, not an official DeepSeek product, and not identical to their internal version or the implementation described in the accompanying paper. It may still contain bugs, its coding style and engineering practices may be inconsistent, and some discussed designs are future directions only. Review and adapt it before using it in your own environment. Licensed MIT.
Compiled from the project README · All rights belong to the original author
Report issue / Request removal