@seekjourney: Everyone may have missed a hidden treasure in DeepSeek Harness's release yesterday: .agents/skills/ directly contains 11 real engineering Skills: Code Review, Pre-push Checks, finding over-engineering, Stacked…
Summary
Introduces the .agents/skills/ directory included in the DeepSeek Harness release, containing 11 real engineering Skills (such as code review, pre-push checks, CoT leakage cleanup, etc.), which is equivalent to open-sourcing part of their engineering standards and Agent SOP.
View Cached Full Text
Cached at: 08/14/26, 03:35 PM
Everyone may have overlooked a hidden gem in yesterday’s DeepSeek Harness release: .agents/skills/ directly contains 11 real-world engineering Skills: Code Review, Pre-push Checks, over-engineering detection, Stacked PR, documentation sync, bilingual translation, CoT leak cleanup, and even auto-recording GIFs for GUI PR acceptance. It looks like DeepSeek has directly open-sourced part of its engineering standards + Review methodology + Quality Gates + Agent SOP. If you’re building a Coding Agent / Harness, these are definitely worth dissecting one by one.
https://github.com/deepseek-ai/deepseek-harness/tree/master/.agents/skills…
deepseek-ai/deepseek-harness
Source: https://github.com/deepseek-ai/deepseek-harness
DeepSeek Harness
English | 中文
DeepSeek Harness (dsh) is an open-source agent harness developed by DeepSeek AI (https://deepseek.com). It uses an architecture where everything is a plugin, and is powered by Cordis (https://github.com/cordiverse/cordis), whose design is described in A Programming Paradigm for Spatiotemporal Composability (https://github.com/cordiverse/paper).
Developer preview
DeepSeek Harness is currently in developer preview and is iterating rapidly. THERE WILL BE COMPATIBILITY-BREAKING CHANGES.
Run
Run from npm
Install Node.js, then run:
sh npx @deepseek-ai/dsh web
The command starts the Web UI, served at http://127.0.0.1:3080 by default. See Web UI guide.
Run from source
To run from a repository checkout:
sh git clone https://github.com/deepseek-ai/deepseek-harness.git cd deepseek-harness pnpm install pnpm run build pnpm dsh web
Community and support
- Feel free to submit feedback or bug reports through GitHub Discussions (https://github.com/deepseek-ai/deepseek-harness/discussions).
- Add the
dsh-plugin(https://github.com/topics/dsh-plugin) topic to your plugin repository for discoverability. - Join DeepSeek Harness Discord community.
Contributing
See CONTRIBUTING.md.
Development
Start with the development guide and architecture documentation. For agents, follow AGENTS.md.
License
MIT Third-party dependencies and their licenses are disclosed in THIRD_PARTY_NOTICES.md.
Similar Articles
@Saccc_c: Deepseek harness's trace mode is so interesting, now you can completely know the progress of Agent's work, and your Agent employees will never dare to slack off again. Its specific thinking steps, tool calls, and tool return results are all clearly visible. The most interesting thing is that it uses different colors to distinguish when AI executes tasks…
DeepSeek Harness has been released, featuring a trace mode that visualizes Agent's thinking steps, tool calls, and return results, and uses colors to distinguish different types of content. It can be installed via npm and is suitable for developers to observe Agent work progress.
@Saccc_c: L站 netizens are still too impressive. Someone reverse-engineered part of the technical architecture based on the leaked DeepSeek Harness closed-beta ecosystem plugin repository on GitHub, for reference. All I can say is I'm looking forward to DeepSeek Harness!
L站 netizens reverse-engineered part of the technical architecture based on the leaked DeepSeek Harness closed-beta ecosystem plugin repository on GitHub, sparking anticipation for DeepSeek Harness.
@jakevin7: DeepSeek Flash is indeed powerful—come feel this long-horizon capability. Agentic abilities are now far stronger. It even discovered the agent swarm tool call in the harness on its own and handled the splitting and orchestration well. This wasn't in the prompt; it discovered it by itself…
The author praises DeepSeek Flash's greatly enhanced long-horizon and agentic abilities, which can automatically discover and combine subagent swarm tool calls in the harness.
@baboonAI4S: https://x.com/baboonAI4S/status/2083921996967227637
A systematic review of 257 open-source projects in the comments of the DeepSeek Harness beta recruitment post reveals a three-tier ecosystem, with DeepSeek support at 49%, and offers recommendations for beta testing priorities.
@tianyi: If you are a developer of an open-source project related to Agent Harness and wish to participate in the beta testing of DeepSeek Harness, you can reply or message me privately. Please include your GitHub ID and representative open-source works.
A DeepSeek team member invites developers of open-source projects related to Agent Harness to participate in the beta testing of DeepSeek Harness, with a GitHub ID and representative works required.